Skip to main content
Download PDF

Proteomic networks and related genetic variants associated with smoking and chronic obstructive pulmonary disease

BMC Genomics volume 25, Article number: 825 (2024) Cite this article

Abstract

Background

Studies have identified individual blood biomarkers associated with chronic obstructive pulmonary disease (COPD) and related phenotypes. However, complex diseases such as COPD typically involve changes in multiple molecules with interconnections that may not be captured when considering single molecular features.

Methods

Leveraging proteomic data from 3,173 COPDGene Non-Hispanic White (NHW) and African American (AA) participants, we applied sparse multiple canonical correlation network analysis (SmCCNet) to 4,776 proteins assayed on the SomaScan v4.0 platform to derive sparse networks of proteins associated with current vs. former smoking status, airflow obstruction, and emphysema quantitated from high-resolution computed tomography scans. We then used NetSHy, a dimension reduction technique leveraging network topology, to produce summary scores of each proteomic network, referred to as NetSHy scores. We next performed a genome-wide association study (GWAS) to identify variants associated with the NetSHy scores, or network quantitative trait loci (nQTLs). Finally, we evaluated the replicability of the networks in an independent cohort, SPIROMICS.

Results

We identified networks of 13 to 104 proteins for each phenotype and exposure in NHW and AA, and the derived NetSHy scores significantly associated with the variable of interests. Networks included known (sRAGE, ALPP, MIP1) and novel molecules (CA10, CPB1, HIS3, PXDN) and interactions involved in COPD pathogenesis. We observed 7 nQTL loci associated with NetSHy scores, 4 of which remained after conditional analysis. Networks for smoking status and emphysema, but not airflow obstruction, demonstrated a high degree of replicability across race groups and cohorts.

Conclusions

In this work, we apply state-of-the-art molecular network generation and summarization approaches to proteomic data from COPDGene participants to uncover protein networks associated with COPD phenotypes. We further identify genetic associations with networks. This work discovers protein networks containing known and novel proteins and protein interactions associated with clinically relevant COPD phenotypes across race groups and cohorts.

Peer Review reports

Introduction

In the US, chronic obstructive pulmonary disease (COPD) is a major public health concern as the fourth leading cause of death [1], affecting more than 16 million adults [2]. COPD is characterized by lung inflammation and the diagnosis of chronic airflow obstruction is made using spirometry [3]. Tobacco smoking is the primary exposure risk factor for the development of COPD in the US. Staudt et al. [4] showed that tobacco smoke diminished the capacity to regenerate airway epithelium in COPD. It is not unexpected, therefore, that 42.3% of current and former smokers with normal spirometry [5] have respiratory symptoms and evidence of emphysema or airway thickening on chest computed tomography (CT) scans.

Forced expiratory volume in one second (FEV1) and percent emphysema (%LAA950) are clinically observable characteristics related to symptoms, exacerbations, and response to treatment [6]. Being a non-invasive, inexpensive, highly accessible, and easily reproducible method, spirometry is the current gold standard for diagnosing and monitoring COPD progression [7]. Emphysema is another phenotype of COPD, which describes obliteration of the acinar units of the lung [8]. Emphysema can be quantified by lung density measured from CT images in which dense lung tissue is replaced by less dense air [9].

Recent advances in high throughput technologies allow investigators to collect data from multiple biological layers including the genome, transcriptome, and metabolome [10,11,12]. In particular, the proteome, where peptide and protein abundance are quantified, has posed a great advantage in studying complex diseases such as COPD since proteins play direct functional roles in biological systems and may provide more relevant information related to disease mechanisms than transcriptional profiling [13]. Previous studies have focused on individual proteins associated with COPD [14]. Lee et al. [15] identified eight up-regulated proteins in the COPD group in comparison with the nonsmoker group. Similarly, Ohlmeier et al. [16] observed increased levels of surfactant protein A (SP-A) in COPD participants but not in the normal or fibrotic lung by investigating changes in the proteome from human lung tissue.

While these studies identified individual protein biomarkers with prognostic potential, they were limited by small sample sizes in hard-to-obtain lung tissue and lack the additional predictive power gained by simultaneously considering a collection of related biomarkers and their interactions [17]. Consequently, network-based analyses have emerged as a powerful framework to characterize changes in multiple molecular entities and their interconnections that may not be captured by single molecular features [18]. Obeidat et al. [19] constructed networks of co-expressed genes from peripheral blood of COPD patients using weighted correlation network analysis (WGCNA) [20] and identified networks associated with FEV1 and enriched in interleukin (IL)-10 and IL-8 signaling pathways. In another study, Mammen et al. [21] performed network analysis on proteomic data collected from bronchoalveolar lavage of the epithelial lining fluid (BALF) samples and identified 233 differentially expressed proteins in moderate COPD compared to controls. Topological analysis of these proteins suggested the importance of intercellular adhesion molecule 1 (ICAM1), galectin-3, fibronectin, and vimentin in mediating inflammation and fibrogenesis.

Most large-scale omics studies for COPD have been conducted in primarily European ancestry populations while only a limited number of relatively smaller-sized studies have focused on other populations [15, 22,23,24,25,26]. Polygenic risk scores (PRS) provide complementary information for predicting COPD and related phenotypes [27], however, they present a large amount of uncertainty which limits the transferability across ancestry groups [28, 29]. Motivated by the lack of COPD omics studies in non-European ancestries, we conducted proteomic analyses on a large cohort that includes > 1,500 self-described African American (AA) subjects to gain more insights into potential proteomic signatures associated with the disease. We leverage proteomic data and network-based approaches to identify protein networks associated with COPD phenotypes separately in AA and Non-Hispanic White (NHW) participants.

In this work, we used sparse multiple canonical correlation network analysis (SmCCNet) [30] to construct proteomic networks associated with two COPD phenotypes (FEV1 and emphysema quantified as percentage of low-attenuation areas defined by voxels with Hounsfield Units < 950 (%LAA950)) and a relevant exposure (current smoking status) across two race groups (AA and NHW).

The resulting networks were compared to identify common, phenotype- and race-group-specific proteins and their corresponding interactions to gain insights into the underlying mechanisms of COPD. As proteins can have strong genetic associations [31,32,33] that may reflect upstream regulatory processes, we also performed a genome-wide quantitative trait loci (QTL) analysis to identify loci associated with each network, in addition to QTL analyses of individual proteins in the networks. Through colocalization and conditional analyses, we further investigated whether the genetic associations observed were due to individual effects of the proteins in the network versus a cumulative effect of the network. Finally, we demonstrated that networks for smoking and %LAA950 built in one race group generally transfer to the other and that networks also validated in an external cohort, the SubPopulations and InteRmediate Outcome Measures in COPD Study (SPIROMICS).

Materials and methods

COPD cohorts

COPDGene [34, 35] (Clinical Trial Registration NCT02445183, https://clinicaltrials.gov/study/NCT02445183 (2 March 2021, accessed 29 August 2023) is a large, multi-center observational study that enrolled 10,198 current and former smokers with at least a 10 pack-year history of smoking, as well as additional never smoker controls (< 100 lifetime cigarettes) with and without COPD, 45–80 years old, with 2/3 non-Hispanic White and 1/3 African Americans. Genotyping data were from the enrollment visit. Proteomics was generated at the five-year follow up using the SomaScan v4.0 platform (2013 and 2017, Visit 2) [34] [36] (Supplement Fig. 1). All study participants provided informed written consent.

SPIROMICS [37] (Clinical Trial Registration NCT01969344, https://clinicaltrials.gov/study/NCT01969344 (11 January 2023, accessed 29 August 2023) is a multi-center observational study that enrolled 2,973 current and former smokers with at least 20 pack-years of smoking between November 2011 to January 2015. Subject were between 40–80 years of age at the time of enrollment and were categorized into never-smokers (< 1 pack-year, Stratum 1) or history of smoking (> 20 pack years) and divided by spirometry into strata; Stratum 2: FEV1/FVC > 0.7 and FVC > LLN; Stratum 3 : FEV1/FVC < 0.07 and FEV1 > 50% predicted; Stratum 4: FEV1/FVC < 0.07 and FEV1<50% predicted). The cohort is multiracial with 73% non-Hispanic White, 18% African American, and 9% other races. Fasting blood was collected at visit 1 in vacutainer EDTA plasma tube, immediately spun, aliquoted, frozen, and stored at − 80°C [38]. For replication we used the smokers (strata 2–4) non-Hispanic White and African American race groups at Visit 1 who had SomaScan v 4.1 profiles (n = 1792). All study participants provided informed written consent (Supplement Table 2, Supplement Fig. 2).

COPDGene cohort demographics

Proteomic analyses included 3,173 COPDGene participants. Demographics and relevant clinical characteristics of participants, stratified by self-identified race, are shown in Table 1. All participants are current or former smokers. We applied a matching approach in an attempt to better match the NHW and AA groups in terms of age, smoking status, sex, and GOLD stage (see Supplement Table 1 and Table 1 for details). Further details on matching are in Supplementary Methods.

Table 1 Characteristics of COPDGene and SPIROMICS by race group
Full size table

COPD phenotypes and exposures

COPD was defined by spirometric evidence of airflow obstruction, which was computed as a ratio of post-bronchodilator FEV1 to forced vital capacity (FVC). The Global Obstructive Lung Disease (GOLD) system is used to grade COPD: in our smoking groups (current and former) GOLD 0 represents an individual without COPD (FEV1 > 80%; FEV1/FVC ≥ 0.7), GOLD 1 (FEV1 ≥ 80%; FEV1/FVC < 0.7), GOLD 2 (50% ≤ FEV1 < 80%; FEV1/FVC < 0.7), GOLD 3 (30% ≤FEV1 < 50%; FEV1/FVC < 0.7), and GOLD 4 (FEV1 < 30%; FEV1/FVC < 0.7), respectively represent the smoker control, mild, moderate, severe, and very severe stages of COPD. Individuals with an FEV1/FVC ≥ 0.70 and FEV1% predicted ≤ 80% were defined as having Preserved Ratio Impaired Spirometry (PRISm) [39]. We use FEV1 as measured in liters as opposed to the race-based percent predicted which can create bias, but adjust for other covariates described below. Emphysema was captured as the log-transformed percentage of lung voxels with Hounsfield Units (HU) < − 950 (%LAA950) on chest CT scan. This metric is also called percentage of low attenuation areas (%LAA). Current smoking status was defined as “former smokers” if they had not smoked any cigarettes within the last 30 days or “current smokers” if they had. Data to calculate the number of pack-years a person smoked were self-reported and calculated based on the packs of cigarettes smoked per-day multiplied by the total number of years smoked.

Matched non-Hispanic White and African American race groups

COPDGene non-Hispanic White (NHW) and African American (AA) groups had different sample sizes as well as key characteristics such as age, current smoking status, sex, and severity of COPD (GOLD Stage). Therefore, we applied a matching approach using SAS version 9.4 SAS/STAT version 15.1, surveyselect procedure to better match groups on these variables, with a particular focus on current smoking and GOLD stage. Details are provided in Supplementary Methods.

Proteomic platforms and final data sets

Plasma protein levels were quantified with SomaScan and quality controlled by SomaLogic (Boulder, Co) [40]. Further details on SomaScan platforms are provided in Supplementary Methods. For COPDGene, the final matched race groups were 1,660 NHW and 1,513 AAs (Table 1, Supplementary Methods). For SPIROMICS, the final replication group was 1,792 subjects (1,459 NHW and 333 AA) (Table 1).

Covariate adjustment

To account for potential confounding effects, we adjusted proteomic data for sex, age, and clinical center. Specifically, we fit an ordinary least squared regression model for each protein such that its abundance was used as the response variable and the three variables (sex, age, and clinical center) as covariates. The resulting residuals were used as input for downstream analysis.

Network analysis

Network construction

We used SmCCNet [30] to generate protein subnetworks associated with each COPD phenotype (FEV1 and %LAA950) and smoking (Supplement Fig. 3). SmCCNet was originally developed to consider multiple omics data sets, so we modified the SmCCNet algorithm to a single omics setting by removing scaling between pairs of omics data. This proposed method has two implementations: one for continuous outcomes (applied to %LAA950 and FEV1) and one for binary outcomes (applied to smoking status). The continuous outcome scenario follows the SmCCA framework and implements sparse canonical correlation analysis. The binary exposure scenario implements sparse partial least square discriminant analysis (SPLS-DA) [41, 42], by performing a classification task under a supervised setting with a two-stage procedure. For the first step, the projection matrix is extracted with regular partial least square assuming a continuous phenotype. For the second step, the projected data is used to fit a logistic regression model. Details are provided in Supplementary Methods.

Network trimming and summarization

The subnetworks obtained through hierarchical clustering may still contain some proteins which are not strongly associated with the phenotype of interest. Therefore, our next step was to further trim the subnetworks such that only the most informative proteins were retained using the PageRank algorithm [43]. We then summarized each subnetwork using the NetSHy approach which applied principal component analysis (PCA) on the combination of both protein abundance and topological properties to obtain the first three low-dimensional summarization scores, referred to as NetSHy scores [44]. In all but one case noted in the Results, the top three scores accounted for over 40% of the cumulative variance explained. We calculated the correlation between each NetSHy score with the corresponding phenotype. Recall that each NetSHy score is a weighted average abundance of all proteins in the network with the relative weights determined by the corresponding loadings. By ranking absolute values of the loadings, we can identify the top five proteins that contribute the most to each NetSHy score in each network. We denote these as the top five loading proteins. We use the L2-norm explained, defined as the sum of squares of the top five protein’s loadings from each NetSHy PC, to check the total contribution of these proteins to their corresponding NetSHy PC. We found that among all 18 NetSHy PCs (six networks X three PCs), 15 of them have at least 90% of the L2-norm explained, and all of them have at least 65% L2-norm explained by the top five proteins.

Based on the topology of each network, we compute the total connection strength of each protein by adding up all the edges connecting that protein to every other protein in the network. Network density refers to the ratio of the number of actual connections observed in a network to the total number of possible connections in that network. We define hub proteins as those proteins that have the top five largest total connection strength values (in some cases there are ties, see Supplementary Table 3). We use a ranking approach, as opposed to absolute cutoffs for the number of connections, as the density of the networks may vary.

Statistical test for comparing subnetworks

We quantify the similarities and differences between subnetworks associated with each phenotype and exposure across the two race groups using the p-norm difference test (PND) with the exponent p = 6, referred to as PND6, which was shown to be a top performing test by Arbet et al. [45]. For each phenotype and exposure, we compute a PND6 statistic which aggregates all the edge-wise differences across the two group-specific subnetwork adjacency matrices. Using a non-parametric permutation method, we derive the sampling distribution under the null hypothesis to generate the corresponding p-values. In our setup, p-values that are smaller than a significance level α correspond to rejecting the null hypothesis at the α level, indicating that the two comparing subnetworks are different. More details are provided in Supplementary Methods.

Network projection

In addition to a direct subnetwork comparison using the PND6 test statistic, we also investigate the similarities and differences between race-specific subnetworks by projecting a subnetwork derived from one race group onto another and vice versa. Specifically, we impose the subnetwork connectivity from one group onto the proteomic data of the other group to compute NetSHy scores as in [44], referred to as projection scores. We calculate correlations between these scores with each respective phenotype or exposure to statistically compare with the original correlations. This procedure is also used to compare subnetworks between COPDGene and SPIROMICS cohorts. Details are provided in Supplementary Methods.

Network quantitative trait loci (nQTL) analysis

COPDGene WGS data was generated by the NHLBI Trans-Omics for Precision Medicine (TOPMed) program [46]. Details are provided in Supplementary Methods. For each subnetwork, we performed a genome-wide network quantitative trait locus (nQTL) analysis of the 3 inverse-normalized NetSHy scores (NetSHy1, NetSHy2, NetSHy3) assuming an additive model for genotype [44]. We regressed the NetSHy scores on each genetic variant separately adjusting for covariates depending on the phenotype used to generate the sub-network. For FEV1 and %LAA950 – the nQTL model was adjusted for sex, age, BMI, smoking status, and 6 genetic PCs to adjust for genetic similarity. For smoking - the nQTL model was adjusted for sex, age, BMI, and 6 genetic PCs [47]. We conducted nQTL analysis on the University of Michigan Encore [48] server’s “Efficient and parallelizable association container toolbox” (EPACTS) [49]. Briefly, EPACTS efficiently performs statistical tests between phenotypes/exposure and sequence data through a user-friendly interface.

Conditional nQTL analysis

As a secondary analysis, we conducted genome-wide association tests for top proteins contributing to each NetSHy score, defined by their contribution to the NetSHy score. We regressed the inverse-normalized protein levels adjusting for covariates in the same manner as for the NetSHy network scores. If associations for phenotype and protein were observed in the same chromosomal locus, colocalization analysis was performed to assess whether the same genetic region contributed to both the genetic associations. If colocalization was observed, genome-wide analysis of phenotype was rerun with normalized protein value as an additional covariate, testing the hypothesis that the network quantitative trait loci (nQTLs) were driven by single protein quantitative trait loci (pQTLs). Further details are described in the Supplementary Methods.

Pathway overrepresentation enrichment meta-analysis

Proteins from each network were input into Metascape [50] as discrete lists. Uniprot identifiers were mapped to Entrez gene IDs. These genes were then assessed for enrichment in a variety of databases (Functional Set: Gene Ontology (GO): Molecular Functions; Pathway: GO: Biological Processes, Hallmark, Reactome, KEGG Pathway, WikiPathways, Canonical Pathways, BioCarta Gene Sets, PANTHER Pathway; Structural Complex: GO: Cell Components, CORUM). All proteins assayed by the SomaScan v4.0 platform were included as a background list for enrichment. Protein-protein interaction (PPI) networks obtained from STRING [51], BioGrid [52], OmniPath [53], and InWeb_IM [54] were additionally seeded with these genes and the MCODE algorithm [55] was used to identify subnetworks of connected proteins.

Results

Despite matching some differences between NHW and AA still exist, but these differences are not clinically large. The biggest differences seen are with COPD Gold stages with AA having a larger percentage with normal lung function and a lower median number of pack-years of smoking. In the SPIROMICS cohort, which was not matched there are large differences in age, sex, smoking status, and severity of COPD. Both cohort’s AA population had higher levels of emphysema (Table 1). While the two cohorts are COPD cohorts, their recruitment criteria were different, and therefore there are differences in their overall characteristics with SPIROMICS being on average older, with a higher percentage of NHW, males, current smokers with a higher number of pack-years, more severe COPD and emphysema (Supplement Table 2).

Protein networks associated with COPD phenotypes and smoking exposure

Smoking

The NHW smoking network consisted of 34 proteins while the AA smoking network consisted of 17 proteins (Fig. 1). Of those network proteins, only 27 and 7 proteins for NHW and AA respectively were significant in the univariate analysis at FDR < 0.10 (Table 2, Supplement Table 3, Supplement Figs. 5 and 6). Across the two race groups, there were seven overlapping proteins including UCRP, PAP1, LPLC1, IGFBP-1, alkaline phosphatase placental type (ALPP), leptin, and EDIL3. In the NHW network, correlation between each protein and smoking status ranged from − 0.20 to 0.36. The range of correlation between the proteomic data and smoking status was smaller in the AA network (-0.17 to 0.23). Correlations between networks in NHW and AA groups with the smoking exposure were 0.33 and 0.23, respectively (Table 2). Both networks displayed high connectivity such that each node was connected to every other node, leading to a corresponding network density equal to one. In both networks, ALPP had many heavily weighted connections. In particular, the connection strengths from ALPP to leptin, CRLD2, and GKN2 were 1, 0.79, and 0.76, respectively, in the NHW network. Similarly, in the AA network, ALPP was strongly connected to EDIL3 (1.0), leptin (0.9), and IGFBP-1 (0.67). As expected, by intersecting the lists of hub proteins and top loading proteins, we observed that hub proteins generally contributed more to the network summary scores than other proteins across the two race groups. For instance, in the NHW network, hub proteins such as ALPP, leptin, and PPBN were also among those with the largest loadings. Similarly, hub nodes in the AA network including ALPP, leptin, and trypsin-2 also contributed the most to the network summary score.

Fig. 1
figure 1

Networks associated with smoking exposure for NHW and AA race groups. NHW network consists of 34 proteins while AA network has 17 proteins. Red edges connect proteins that have positive correlations with smoking exposure while blue edges link negatively correlated proteins. The line width is proportional to the connection strength. Correlations between networks in NHW and AA race groups. with the smoking exposure are 0.33 and 0.23, respectively. Dashed boxes indicate overlapping proteins between the two networks

Full size image
Table 2 Summary results for each phenotype- and group-specific subnetwork
Full size table

We used a statistical approach to compare the adjacency matrices representing the two race-specific networks. Given that the two networks had different sizes (34 vs. 17 proteins), we found a union set of 44 proteins present in either or both networks, prior to calculating the p-norm difference test with exponent equal to 6 (PND6) (See Methods). Table 3 shows the resulting test statistics and p-values when comparing smoking-associated networks to indicate that networks associated with smoking are similar across NHW and AA race groups (PND6 = 0.340, p-value = 0.955). Supplement Fig. 4a displays the corresponding heatmap for edge-wise differences in networks associated with smoking exposure between NHW and AA groups. In alignment with the PND6 test, we observed more white or lighter red areas, highlighting the similarity of smoking-associated networks across the two race groups. Additionally, Table 4 summarizes the similarities and differences between smoking-associated subnetworks by projecting a subnetwork across race groups and/or cohorts, which is a complementary approach that does not require adjacency matrices in each group. Within COPDGene, we computed the cross-race correlations by projecting the NHW subnetwork onto AA data and vice versa, and we observed similar correlations across the two race groups. Specifically, when the AA subnetwork (C-AA) was projected to NHW proteomic (C-NHW) data, the first two projection correlations were 0.354 and 0.125, respectively. The original correlations, 0.329 and 0.208, fell within the corresponding 95% bootstrap confidence intervals (CIs) of (0.303, 0.403) and (0.057, 0.202), respectively. Similarly, when we projected the NHW subnetwork (C-NHW) onto AA (C-AA) data, the corresponding 95% CIs also captured the observed correlations, demonstrating the similarity between subnetworks across the two race groups within the same cohort.

Table 3 Results of direct network comparison using subnetworks associated with smoking exposure, FEV1, and %LAA950 across NHW and AA 
Full size table
Table 4 Summary results of network projections using subnetworks associated with smoking exposure across NHW and AA, in two independent cohort studies COPDGene (C) and SPIROMICS (S)
Full size table

We further projected the subnetworks derived from COPDGene (C) onto the data in SPIROMICS (S) to assess the replicability of the subnetworks across independent cohorts. By projecting the NHW subnetwork derived from COPDGene (C-NHW), onto the NHW data in SPIROMICS (S-NHW) we obtained the first two cross-cohort correlations of 0.373 and 0.186, respectively. Note that the 95% CI of the first projection component (0.334, 0.416) was significantly higher than the original correlation of 0.329. Similarly, when we projected the COPDGene AA subnetwork (C-AA) onto SPIROMICS NHW (S-NHW) data, the first projection correlation was 0.393 and its 95% CI was (0.347, 0.432). Once again, the confidence interval was higher than the original correlation of 0.329. Such consistent projection correlations indicate a high level of replicability of the subnetworks associated with smoking exposure across independent cohorts. In a similar manner, we projected the C-AA subnetwork onto the SPIROMICS AA (S-AA) data and also observed similar results (Table 4). In summary, these results provide further evidence of the replicability of the smoking subnetworks across cohorts, even when considering different race groups.

FEV1

There were 13 and 22 proteins present in the NHW and AA networks for FEV1, respectively, with sRAGE present in both networks (Fig. 2). Of those network proteins, only two and one protein(s) for NHW and AA respectively were significant in the univariate analysis at FDR < 0.10 (Table 2, Supplement Table 3, Supplement Figs. 7 and 8). In the AA network, sRAGE was strongly connected to carboxypeptidase B (1.0) and EDIL3 (0.53) while displaying relatively weaker relationships (< 0.33) with the remaining nodes. In the NHW network, sRAGE showed strong connections to renin (0.93) and lefty-A (0.7) while maintaining moderate relationships of at least 0.5 to other proteins. Correlations between individual proteins with FEV1 ranged from − 0.11 to 0.1 in the NHW network and from − 0.09 to 0.12 in the AA network. Correlations between NetSHy1 of networks derived from NHW and AA participants with FEV1 were 0.13 and 0.14, respectively (Table 2).

Fig. 2
figure 2

Networks associated with FEV1 for NHW and AA race groups. NHW network consists of 13 proteins while AA network has 22 proteins. Red edges connect proteins that have positive correlations with smoking exposure while blue edges link negatively correlated proteins. The line width is proportional to the connection strength. Correlations between networks in NHW and AA race groups with the smoking exposure are 0.13 and 0.14. Dashed boxes indicate overlapping proteins between the two networks

Full size image

We next investigated potential overlap between the NHW and AA networks. Using the PND6 method, we found a significant difference between the two networks (p-value < 0.001, Table 3, Supplement Fig. 4b). The projection approach also showed poor performance, suggesting notable differences between the FEV1 networks across the two race groups. We further projected the subnetworks derived from COPDGene (C) onto the data in SPIROMICS (S) to assess the replicability of the subnetworks across independent cohorts. By projecting the C-NHW subnetwork onto the S-NHW data and vice versa, we found that the corresponding 95% CIs also captured the original correlations, suggesting some degree of replicability across cohorts for the same race group (Supplement Table 4a). However, the CIs were relatively wider than with smoking, which might be due to more variation in the subnetworks associated with this phenotype. These observations indicate some moderate degree of transferability of FEV1 associated networks across cohorts for the same race group. However, the results also highlight potential variations in the subnetworks associated with FEV1 across race groups, emphasizing the importance of considering group-specific characteristics when studying this phenotype.

%LAA950

There were 21 and 104 proteins present in NHW and AA networks for %LAA950, respectively (Fig. 3). Of those network proteins, only four and six proteins for NHW and AA respectively were significant in the univariate analysis at FDR < 0.10 (Table 2, Supplement Table 3, Supplement Figs. 9 and 10). The AA network is notably larger and denser, and was the only network where the top three summarization scores explained less than 40% of the variability (23% variability explained). Despite this difference there were many consistencies. The two networks had seven proteins in common: PXDN, DAN, FSH, sRAGE, glucagon, SIRB1, RNase 1, and leptin. In the NHW network, the range of correlations between each protein and %LAA950 was between − 0.12 and 0.09, which was similar to that in the AA race group. Correlations between networks derived from NHW and AA groups with %LAA950 were 0.14 and 0.12, respectively (Table 2). Like smoking, the two networks associated with %LAA950 are similar across NHW and AA groups (Table 3, Supplement Fig. 4c). This was also consistent with the projection analysis (Supplement Table 4b) where we found notable similarities between subnetworks associated with %LAA950 across the two race groups within the same cohorts. Furthermore, when comparing the subnetworks associated with %LAA950 across independent cohorts, we also observed consistency in the projections (Supplement Table 4b).

Fig. 3
figure 3

Networks associated with %LAA950 for NHW and AA race groups. NHW network consists of 21 proteins while AA network has 104 proteins. Red edges connect proteins that have positive correlations with smoking exposure while blue edges link negatively correlated proteins. The line width is proportional to the connection strength. Correlations between networks in NHW and AA race groups with the smoking exposure are 0.14 and 0.12. Dashed boxes indicate overlapping proteins between the two networks

Full size image

Enrichment

We performed enrichment of individual proteins within networks and meta-analysis across networks through MetaScape. Significantly enriched pathways are shown in Fig. 4. Top shared pathways identified through meta-analysis include response to hormone (enriched in all gene lists), and regulation of cell activation and response to bacterium enriched in five gene lists (Fig. 4a). Many additional pathways were enriched in multiple gene lists in meta-analysis. Individual enrichment analysis also showed gene lists were enriched for many disease-relevant pathways. For example, in addition to observing many proteins in networks associated with inflammatory and antimicrobial processes, we observe VEGFA-VEGFR2 signaling enriched in %LAA950 NHW, FEV1 NHW, and FEV1 AA networks (Fig. 4b).

Fig. 4
figure 4

Pathway enrichment analysis. ​A. Heatmap of top 20 significantly enriched pathways across protein lists identified through Metascape meta-analysis. Enriched pathways are colored by –log10 (p-value). ​B. Top 10 enriched pathways by network​

Full size image

Network QTLs (nQTLs) show genetic underpinnings of COPD protein networks

We tested for association between the top three NetSHy scores of each protein network and common genetic variants from WGS. Seven NetSHy scores were associated with at least one variant at a genome wide significant level (Table 5, Fig. 5). NetSHy1 of smoking in both AA and NHW participants show genetic association signals on 2q37.1 within or near the gene ALPG. NetSHy2 of FEV1 in NHW participants is associated with variants on chr1 near LEFTY1. NetSHy2 of %LAA950 in AA participants is associated with a single variant in MGAT5, and NetSHy2 and NetSHy3 of %LAA950 in NHW participants show associations with the ABO locus. NetSHy3 of %LAA950 in NHW additionally shows an association signal on chr19 within the gene SIGLEC9. Both ABO lead variants have previously been found to be associated with lung function. Rs8176693 was nominally associated with FEV1/FVC in a European population [56] and rs9921085 is associated with both FEV1(p-value = 1.00 x 10−14) and FVC (p-value = 1.10 x 10 −14) in the UK Biobank [57].

Fig. 5
figure 5

Network QTL associations with protein network NetSHy summary scores. Manhattan plots display the significance (-log10 of p-values, y axis) of genome-wide association tests across all chromosomes (x axis) in African American (left panels A-D), and non-Hispanic White participants (right panels, A-D). Each row corresponds to the following associations: A) Emphysema NetSHy2, B) Emphysema NetSHy3, C) FEV1 NetSHy2, D) Smoking NetSHy1. The top SNP of each association signal is highlighted in red, and it is labeled with the rsID of the SNP, along with the name of the closest gene, when applicable

Full size image
Table 5 Results of genome-wide significant association tests of NetSHy scores
Full size table

We next assessed whether these genetic associations were driven by top proteins in networks. For each NetSHy score with a significant association, we ran a genome-wide association scan for the top five loading proteins contributing to each NetSHy score. We identified associations with proteins in %LAA950 NHW NetSHy 2 (ganglioside GM2 activator), %LAA950 NHW NetSHy3 (cadherin 17 and sRAGE), FEV1 NHW NetSHy2 (regenerating islet derived protein 3 alpha), Smoking AA NetSHy1 (cob(I)yrinic acid a,c-diamide adenosyltransferase mitochondrial, alkaline phosphatase placental type, and insulin growth factor binding protein 1), and Smoking NHW NetShy1 (gastrokine 2, interleukin 12 subunit beta, alkaline phosphatase placental type, and alkaline phosphatase placental like 1) (Supplement Table 5).

In each instance where an nQTL and a single-protein genetic association were on the same chromosome, we tested for colocalization of these signals using coloc. When single protein and nQTL signals colocalized, we reran the associated GWAS with the single protein abundance values included as a covariate to serve as a conditional analysis. After conditional analysis, four NetSHy associations with genetic loci of the seven remain: NHW %LAA950 NetSHy3 – SIGLEC9, AA %LAA950 NetSHy2 - MGAT5, NHW %LAA950 NetSHy2 – ABO, and NHW FEV1 NetSHy2 – LEFTY1 (Supplement Table 5).

Discussion

Summary

We used SmCCNet to generate protein correlation networks associated with FEV1, %LAA950, and smoking status separately in NHW and AA COPDgene participants, containing 13 to 104 proteins. We used smoking exposure as a paradigm to develop methods and contrast race groups as smoking has been well studied. We then used the same approach to investigate other COPD phenotypes such as FEV1 and %LAA950, where our understanding was comparatively limited. The derived networks demonstrated stronger or as strong correlations with phenotypes and exposure than individual proteins demonstrating the benefits of a network approach. Smoking and %LAA950 networks were similar between NHW and AA, and replicated well in the SPIROMICS cohort, while FEV1 networks showed notable differences across the two groups and lower level of replicability.

We ran genome-wide association study analysis on NetSHy scores to identify potential genetic variants associated with the protein networks, which we refer to as nQTLs. Finally, we assessed whether discovered nQTLs were independent of genetic association signals of single top proteins included in the network and identified three genetic variants associated with %LAA950 networks. Through this work, we have identified novel networks of correlated proteins related to COPD phenotypes of interest, as well as common genetic variants associated with these networks. It is worth noting that at many of the proteins in the identified networks were not significantly correlated with the respective phenotype/exposure (Table 2). This demonstrates the advantages of a network approach, which enabled the identification of proteins that were not identified on their own but appear to play a supplementary role in influencing the outcome of interest through their interactions with other proteins that do have a strong association with the phenotype/exposure.

Enrichment analysis of networks demonstrates that network proteins across the phenotypes are associated with processes and pathways such as response to bacterium and antimicrobial peptides, hormone activity, extracellular matrix signaling, and interferon signaling. Antimicrobial proteins include UCRP and LPLC1 in our smoking networks, as well as proteins such as MIP1a and IgD in FEV1 networks, and PXDN and RNase1 in %LAA950 networks. UCRP is integral to the response to infection of multiple respiratory pathogens, including influenza and SARS-CoV-2 [58, 59]. UCRP has previously been demonstrated to be upregulated at the RNA level in alveolar macrophages from COPD patients with more severe disease (based on GOLD staging) [60]. LPLC1 is thought to be involved in innate immune responses to bacterial infection, including in the lung [61]. LPLC1 has previously been demonstrated to be upregulated in sputum of smokers with and without COPD [62]. Furthermore, protein levels in sputum are correlated with smoking pack-years and spirometric measures of lung function (FEV1 & FEV1/FVC) [63]. MIP-1a is an inducible chemokine that promotes inflammation and monocyte and macrophage recruitment. Gene and protein expression is increased in COPD PBMCs relative to healthy controls [64] as well as in sputum [65]. MIP-1a has also been shown to promote tight junction injury in airway epithelium [64]. IgD is the major antigen receptor type on peripheral B-cells. It induces TNF, IL1B, and IL1RN, in addition to other cytokines [66]. Serum IgD has previously been shown to be increased in COPD subjects [67]. PXDN is a heme-containing peroxidase secreted into extracellular matrix that is involved in extracellular matrix formation. PXDN also directly binds gram-negative bacteria in innate immune response, contributing to lung host defense [68]. RNase 1 is an endonuclease targeting single- and double-stranded RNAs. RNASE1 has previously been seen to be upregulated at the gene expression level in PBMCs from COPD patients compared to those from healthy controls [69].

Networks also contain hormones and proteins involved in hormone signaling. These include leptin and IGFBP-1 in smoking networks, glucagon in %LAA950 networks, and renin in FEV1 networks. Leptin is an adipocyte-derived hormone with pro-inflammatory effects. There is conflicting evidence of altered leptin concentrations in COPD [70,71,72]. Low levels of IGFBP-1 which binds both IGF-1 and -2, can indicate impaired glucose tolerance, vascular disease, and hypertension. IGF and IGFBP concentrations have been shown to be altered in COPD and smoking [73]. Glucagon is a pancreatic hormone involved in glucose metabolism and homeostasis and has been shown to reduce airway hyperresponsiveness [74]. Renin is an endopeptidase secreted by the kidneys that targets angiotensinogen, resulting in elevated blood pressure and vasoconstriction [75]. Upregulation of renin-angiotensin signaling can drive pulmonary fibrosis [76]. Angiotensin II regulates response to lung injury and apoptosis in alveolar epithelium [77] and there is some evidence that angiotensin-converting enzyme inhibitors and related drugs result in reduced exacerbations and mortality in COPD [78, 79].

Networks additionally contain molecules involved in tissue remodeling in COPD [80]. For example, our FEV1 networks contain sRAGE, a soluble receptor that binds advanced glycosylation end products, which accumulate in vascular tissues during aging. COPD patients show lower plasma and serum levels of sRAGE. Additionally, sRAGE levels are associated with emphysema severity and reduced FEV1 [81]. Smoking networks contain molecules such as EDIL3 and CRLD2. EDIL3 (EGF-like repeat and discoidin I-like domain-containing protein 3) is an integrin ligand that promotes adhesion of endothelial cells and is involved in angiogenesis and vascular remodeling. Plasma levels of EDIL3 have been shown to be decreased in COPD patients and associated with increased risk of acute exacerbation [82]. CRLD2 (Cysteine-rich secretory protein LCCL domain-containing 2) [CRISPLD2] is a secreted protein that promotes matrix assembly and modulates airway branching and alveogenesis [83]. Glucocorticoid treatment increases gene and protein expression in airway smooth muscle cells, which in turn regulates cytokine levels [84]. Heterozygous knockout mice display features similar to bronchopulmonary dysplasia [85]. CRLD2 has also been shown to attenuate inflammatory signaling induced by LPS in lung fibroblasts and epithelial cells.

Note that some of the proteins above reached nominal significance (p < 0.001) in a univariate analysis (Supplement Table 3) with the respective exposure/outcome but very few reached statistical significance accounting for multiple testing (FDR < 0.10). This further illustrates the benefits of a network approach for identifying proteins that may not have the strongest univariate signal but may have strong interactions with other proteins related to the exposure/outcome.

nQTL Findings

We identified seven nQTL signals for six unique NetSHy scores. nQTLs may play a role in the regulation of the network as opposed to individual pQTL which may only affect a single protein. As nQTLs may be driven by a single strong pQTL, we examined pQTLs for top network proteins and performed colocalization analysis. Four of the seven nQTLs remained associated after conditional analysis adjusting for protein levels of top network proteins with colocalized pQTL protein values. These signals are a variant (rs72846742) on chr2 with AA %LAA950 NetSHy2, a locus overlapping SIGLEC9 on chr19 with NHW %EMP NetSHy3, variants in the ABO locus with NHW %LAA950 NetSHy 2, and a locus on chr1 with NHW FEV1 NetSHy2. We note that while SIGLEC9 was not one of the top five protein loadings for NHW %LAA950, it is present in the network. rs72846742 has been previously associated with smoking intensity [86] and is within the first intron of MGAT5 (alpha-1,6-mannosylglycoprotein 6-beta-N-acetylglucosaminlytransferase). It has also been shown to be an eQTL for MGAT5 in blood by the eQTLGen consortium [87]. This gene encodes a glycosyltransferase primarily implicated in cancer. A recent study reports MGAT5 genetic variation associated with COPD in a Chinese population [88].

The ABO locus has been extensively studied and variants in this gene have been associated with increased risk of numerous diseases. Despite multiple studies of ABO allele frequencies in COPD, no consistent association with disease or related phenotypes has been reported. The lead variant, within an intron of ABO, has been shown to act as both an eQTL and pQTL for ABO [89, 90] and has been associated with numerous phenotypes generally related to blood traits and cardiovascular disease.

Genes within the chr1 locus associated with NHW FEV1 NetSHy2 include EPHX1, TMEM63A, LEFTY1, LEFTY2, and PYCR2. We note that LEFTY2 encodes left-right determination factor 2, a protein that is within the NHW FEV1 network despite not being a top protein loading on NetSHy2. This protein is a secreted ligand that binds TGF-beta receptors. TGF-beta signaling has been implicated in many aspects of COPD [91]. The lead variant in the locus, rs360060, has been shown to act as an eQTL for TMEM63A, LEFTY1, and EPHX1, and is predicted to most likely affect TMEM63A by the OpenTargets Platform [90]. The variants on chr19 are proximal to or within the gene body of SIGLEC9. Protein levels of SIGLEC9 have been shown to be increased in plasma and neutrophils from COPD patients [92] and one variant, rs2075803, has previously been associated with higher exacerbation frequency and greater emphysema in a small cohort [93]. As many nQTL signals seemed driven by single-protein associations, future applications of this framework may address this through approaches such as regressing pQTL signals from the protein data [94].

Limitations and future work

It is important to note that this work was performed using SomaScan platform data, and although there was replication in an independent cohort for the same platform, our findings may not replicate across other proteomic assays. Furthermore, although SomaScan is one of the most comprehensive proteomic profiling methods, it only captures a subset of the proteome so may be missing proteins in the network. However, our genetic investigation of the FEV1 networks showed signals in loci well-studied in the context of COPD, such as EPHX1 [95, 96], although the EPHX1 protein was not included in the SomaScan panel. This finding suggests that the protein networks and its genetic associated loci can capture biologically meaningful signals involved in COPD, even if they are not directly assayed in our study. In the future as platforms become more comprehensive, we will be able to expand on these networks in addition to incorporating other omics measurements. In addition, our results are subject to sources of noise inherent in these types of studies including the use of blood, as opposed to primary tissue, non-fasting measurements, and differences in medication use.

Across all network results, the respective networks had at most 0.33 correlation with smoking and at most 0.14 with FEV1 or %LAA950. Although the correlations with the two phenotypes may not seem strong, they were still larger than the correlation values observed for individual proteins (maximum correlation found for any protein was 0.12 for both phenotypes and race groups) and consistent with what we have observed in our previous biomarker studies [17, 81].

We decided to analyze NHW and AA participants separately within COPDGene for a variety of reasons. In COPDGene, NHW and AA participants display major differences in terms of demographics and disease severity. We implemented a matching scheme to better match NHW and AA groups on age, GOLD stage, and smoking status. In spite of this, groups still exhibited some differences in demographics and disease. To further address demographic confounding with omics signals, we regressed age, sex, and clinical center from the proteomic data prior to network generation. We decided to only include non-modifiable covariates which are unlikely to be influenced by disease in our regression model. Additionally, matching allowed us to down-sample NHW participants to a sample size closer to the AA group, mitigating differences in results that may have been driven by power/sample size issues, which occurs in many studies where data sets from European ancestry individuals are typically much larger than other groups. Finally, we assessed whether networks derived from one race group replicated in the other group in terms of both network structure and NetSHy scores. We found that the smoking and %LAA950 networks were replicated across the race groups indicating shared interactions, even when all proteins in the network did not overlap. On the other hand, FEV1 did not show strong replication across race groups and/or study cohorts. This is not surprising given that spirometry generally shows a great degree of variability [97, 98]. Consequently, networks associated with FEV1 may capture such inherent variability, potentially reducing their replicability.

Although there were many similarities, we emphasize that any observed differences between race groups are likely the result of biases in sampling and potentially driven by social determinants of health (SDoH); differing results between race groups do not indicate nor support differing biology between these groups. SDoH have been associated with proteomic changes linked to increased inflammation in a variety of diseases [99]. When examining self-rated health (SRH) data, poor SRH scores are linked to a rise in inflammatory plasma proteins such as leptin in CVD populations [100]. This current work demonstrates a link between leptin and COPD. There are few studies examining SDoH and COPD and pose a novel path forward for investigation.

Conclusion

In this work, we constructed protein networks that are related to COPD-relevant phenotypes, namely FEV1 and %LAA950, and the primary exposure of smoking, separately in NHW and AA COPDGene participants. We demonstrated the ability to derive sparse protein networks associated with these phenotypes that replicate both across race sub-groups and across cohort studies. By leveraging NetSHy network summarization scores, we were further able to identify common genetic variants associated with NetSHy scores. This work demonstrated both the utility of a combined proteomic-genetic-network approach to identify novel proteins and their interactions involved in COPD phenotypes.

Availability of data and materials

The SomaScan data supporting the conclusions of this article are available from the data coordinating centers of COPDGene and SPIROMICS respectively. The genomic data is available through TOPMed. The open-source code and reproducible analysis scripts can be accessed at https://github.com/KechrisLab/ProteinNetworks/.

References

  1. Kochanek KD. Mortality in the United States. 2016.

    Google Scholar 

  2. Sullivan J, Pravosud V, Mannino DM, et al. National and State Estimates of COPD morbidity and mortality - United States, 2014–2015. Chronic Obstr Pulm Dis Miami Fla. 2018;5:324–33.

    Google Scholar 

  3. McDonough JE, Yuan R, Suzuki M, et al. Small-airway obstruction and emphysema in chronic obstructive pulmonary disease. N Engl J Med. 2011;365:1567–75.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  4. Staudt MR, Buro-Auriemma LJ, Walters MS, et al. Airway basal stem/progenitor cells have diminished capacity to regenerate airway epithelium in chronic obstructive pulmonary disease. Am J Respir Crit Care Med. 2014;190:955–8.

    Article  PubMed  PubMed Central  Google Scholar 

  5. Regan EA, Lynch DA, Curran-Everett D, et al. Clinical and radiologic disease in smokers with normal spirometry. JAMA Intern Med. 2015;175:1539–49.

    Article  PubMed  PubMed Central  Google Scholar 

  6. Han MK, Agusti A, Calverley PM, et al. Chronic obstructive pulmonary disease phenotypes. Am J Respir Crit Care Med. 2010;182:598–604.

    Article  PubMed  PubMed Central  Google Scholar 

  7. Hoesterey D, Das N, Janssens W, et al. Spirometric indices of early airflow impairment in individuals at risk of developing COPD: spirometry beyond FEV1/FVC. Respir Med. 2019;156:58–68.

    Article  PubMed  PubMed Central  Google Scholar 

  8. Singh D. Small airway disease in patients with chronic obstructive pulmonary disease. Tuberc Respir Dis. 2017;80:317–24.

    Article  Google Scholar 

  9. Subramanian DR, Gupta S, Burggraf D, et al. Emphysema- and airway-dominant COPD phenotypes defined by standardised quantitative computed tomography. Eur Respir J. 2016;48:92–103.

    Article  CAS  PubMed  Google Scholar 

  10. Kim M-S, Pinto SM, Getnet D, et al. A draft map of the human proteome. Nature. 2014;509:575–81.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  11. Roadmap Epigenomics Consortium, Kundaje A, Meuleman W, et al. Integrative analysis of 111 reference human epigenomes. Nature. 2015;518:317–30.

    Article  PubMed Central  Google Scholar 

  12. Shin S-Y, Fauman EB, Petersen A-K, et al. An atlas of genetic influences on human blood metabolites. Nat Genet. 2014;46:543–50.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  13. Wu W, Kaminski N. Chronic lung diseases. WIREs Syst Biol Med. 2009;1:298–308.

    Article  CAS  Google Scholar 

  14. Serban KA, Pratte KA, Bowler RP. Protein biomarkers for COPD outcomes. Chest. 2021;159:2244–53.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  15. Lee EJ, In KH, Kim JH, et al. Proteomic analysis in lung tissue of smokers and COPD patients. Chest. 2009;135:344–52.

    Article  CAS  PubMed  Google Scholar 

  16. Ohlmeier S, Vuolanto M, Toljamo T, et al. Proteomics of human lung tissue identifies surfactant protein A as a marker of chronic obstructive pulmonary disease. J Proteome Res. 2008;7:5125–32.

    Article  CAS  PubMed  Google Scholar 

  17. Zemans RL, Jacobson S, Keene J, et al. Multiple biomarkers predict disease severity, progression and mortality in COPD. Respir Res. 2017;18:117.

    Article  PubMed  PubMed Central  Google Scholar 

  18. Liu Z-P, Wang Y, Zhang X-S, et al. Network-based analysis of complex diseases. IET Syst Biol. 2012;6:22–33.

    Article  PubMed  Google Scholar 

  19. Obeidat M, Nie Y, Chen V, et al. Network-based analysis reveals novel gene signatures in peripheral blood of patients with chronic obstructive pulmonary disease. Respir Res. 2017;18:72.

    Article  PubMed  PubMed Central  Google Scholar 

  20. Langfelder P, Horvath S. WGCNA: an R package for weighted correlation network analysis. BMC Bioinformatics. 2008;9: 559.

    Article  PubMed  PubMed Central  Google Scholar 

  21. Mammen MJ, Tu C, Morris MC, et al. Proteomic network analysis of bronchoalveolar lavage fluid in ex-smokers to discover implicated protein targets and novel drug treatments for chronic obstructive pulmonary disease. Pharm Basel Switz. 2022;15:566.

    CAS  Google Scholar 

  22. Prokić I, Lahousse L, de Vries M, et al. A cross-omics integrative study of metabolic signatures of chronic obstructive pulmonary disease. BMC Pulm Med. 2020;20:193.

    Article  PubMed  PubMed Central  Google Scholar 

  23. Ikram MA, Brusselle G, Ghanbari M, et al. Objectives, design and main findings until 2020 from the Rotterdam Study. Eur J Epidemiol. 2020;35:483–517.

    Article  PubMed  PubMed Central  Google Scholar 

  24. Bos D, Portegies MLP, van der Lugt A, et al. Intracranial carotid artery atherosclerosis and the risk of stroke in whites: the Rotterdam study. JAMA Neurol. 2014;71:405–11.

    Article  PubMed  Google Scholar 

  25. Haq I, Chappell S, Johnson SR, et al. Association of MMP – 12 polymorphisms with severe and very severe COPD: A case control study of MMPs – 1, 9 and 12 in a European population. BMC Med Genet. 2010;11:7.

    Article  PubMed  PubMed Central  Google Scholar 

  26. Liu Y, Liu H, Li C, et al. Proteome profiling of lung tissues in Chronic Obstructive Pulmonary Disease (COPD): platelet and macrophage dysfunction Contribute to the Pathogenesis of COPD. Int J Chron Obstruct Pulmon Dis. 2020;15:973–80.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  27. Moll M, Lutz SM, Ghosh AJ, et al. Relative contributions of family history and a polygenic risk score on COPD and related outcomes: COPDGene and ECLIPSE studies. BMJ Open Respir Res. 2020;7: e000755.

    Article  PubMed  PubMed Central  Google Scholar 

  28. Zhao Z, Fritsche LG, Smith JA, et al. The construction of cross-population polygenic risk scores using transfer learning. Am J Hum Genet. 2022;109:1998–2008.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  29. Ding Y, Hou K, Burch KS, et al. Large uncertainty in individual polygenic risk score estimation impacts PRS-based risk stratification. Nat Genet. 2022;54:30–9.

    Article  CAS  PubMed  Google Scholar 

  30. Shi WJ, Zhuang Y, Russell PH, et al. Unsupervised discovery of phenotype-specific multi-omics networks. Bioinformatics. 2019;35:4336–43.

    Article  PubMed  PubMed Central  Google Scholar 

  31. Sun W, Kechris K, Jacobson S, et al. Common genetic polymorphisms influence blood biomarker measurements in COPD. PLOS Genet. 2016;12:e1006011.

    Article  PubMed  PubMed Central  Google Scholar 

  32. Moll M, Jackson VE, Yu B, et al. A systematic analysis of protein-altering exonic variants in chronic obstructive pulmonary disease. Am J Physiol-Lung Cell Mol Physiol. 2021;321:L130-43.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  33. Shrine N, Izquierdo AG, Chen J, et al. Multi-ancestry genome-wide association analyses improve resolution of genes and pathways influencing lung function and chronic obstructive pulmonary disease risk. Nat Genet. 2023;55:410–22.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  34. Regan EA, Hokanson JE, Murphy JR, et al. Genetic epidemiology of COPD (COPDGene) study design. COPD. 2010;7:32–43.

    Article  PubMed  Google Scholar 

  35. National Jewish Health. COPDGene/Lung Cancer Center Database. Clinical Trial Registration NCT02445183, clinicaltrials.gov. 2021. https://clinicaltrials.gov/study/NCT02445183. Accessed 29 Aug 2023.

  36. Bradford E, Jacobson S, Varasteh J, et al. The value of blood cytokines and chemokines in assessing COPD. Respir Res. 2017;18:180.

    Article  PubMed  PubMed Central  Google Scholar 

  37. University of North Carolina, Chapel Hill. Subpopulations and Intermediate Outcome Measures in COPD Study. Clinical Trial Registration NCT01969344, clinicaltrials.gov. 2023. https://clinicaltrials.gov/study/NCT01969344. Accessed 29 Aug 2023.

  38. Couper D, LaVange LM, Han M, et al. Design of the subpopulations and intermediate outcomes in COPD study (SPIROMICS). Thorax. 2014;69:491–4.

    Article  PubMed  Google Scholar 

  39. Wan ES, Castaldi PJ, Cho MH, et al. Epidemiology, genetics, and subtyping of preserved ratio impaired spirometry (PRISm) in COPDGene. Respir Res. 2014;15:89.

    Article  PubMed  PubMed Central  Google Scholar 

  40. Serban KA, Pratte KA, Strange C, et al. Unique and shared systemic biomarkers for emphysema in Alpha-1 Antitrypsin deficiency and chronic obstructive pulmonary disease. eBioMedicine; 84. Epub ahead of print 1 October 2022. https://doi.org/10.1016/j.ebiom.2022.104262.

  41. Chun H, Keleş S. Sparse partial least squares regression for simultaneous dimension reduction and variable selection. J R Stat Soc Ser B Stat Methodol. 2010;72:3–25.

    Article  Google Scholar 

  42. Chung D, Keles S. Sparse partial least squares classification for high dimensional data. Stat Appl Genet Mol Biol. 2010;9:Article17.

    Article  PubMed  Google Scholar 

  43. Google’s PageRank and Beyond. 2012. https://press.princeton.edu/books/paperback/9780691152660/googles-pagerank-and-beyond. Accessed 9 Jan 2023.

  44. Vu T, Litkowski EM, Liu W, et al. NetSHy: network summarization via a hybrid approach leveraging topological properties. Bioinformatics. 2023;39: btac818.

    Article  CAS  PubMed  Google Scholar 

  45. Arbet J, Zhuang Y, Litkowski E, et al. Comparing statistical tests for differential network analysis of gene modules. Front Genet. 2021;12. https://www.frontiersin.org/articles/10.3389/fgene.2021.630215. Accessed 22 March 2023.

  46. Taliun D, Harris DN, Kessler MD, et al. Sequencing of 53,831 diverse genomes from the NHLBI TOPMed Program. Nature. 2021;590:290–9.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  47. NHLBI Trans-Omics for Precision Medicine WGS-About TOPMed, https://nhlbiwgs.org/. Accessed 1 July 2020.

  48. Encore | Dashboard, https://encore.sph.umich.edu/. Accessed 14 Jan 2021.

  49. EPACTS - Genome Analysis Wiki. https://genome.sph.umich.edu/wiki/EPACTS . Accessed 14 Jan 2021.

  50. Zhou Y, Zhou B, Pache L, et al. Metascape provides a biologist-oriented resource for the analysis of systems-level datasets. Nat Commun. 2019;10:1523.

    Article  PubMed  PubMed Central  Google Scholar 

  51. Szklarczyk D, Franceschini A, Wyder S, et al. STRING v10: protein-protein interaction networks, integrated over the tree of life. Nucleic Acids Res. 2015;43:D447-452.

    Article  CAS  PubMed  Google Scholar 

  52. Stark C, Breitkreutz B-J, Reguly T, et al. BioGRID: a general repository for interaction datasets. Nucleic Acids Res. 2006;34:D535-9.

    Article  CAS  PubMed  Google Scholar 

  53. Türei D, Korcsmáros T, Saez-Rodriguez J. OmniPath: guidelines and gateway for literature-curated signaling pathway resources. Nat Methods. 2016;13:966–7.

    Article  PubMed  Google Scholar 

  54. Li T, Wernersson R, Hansen RB, et al. A scored human protein–protein interaction network to catalyze genomic interpretation. Nat Methods. 2017;14:61–4.

    Article  CAS  PubMed  Google Scholar 

  55. Bader GD, Hogue CW. An automated method for finding molecular complexes in large protein interaction networks. BMC Bioinformatics. 2003;4:2.

    Article  PubMed  PubMed Central  Google Scholar 

  56. Shrine N, Guyatt AL, Erzurumluoglu AM, et al. New genetic signals for lung function highlight pathways and chronic obstructive pulmonary disease associations across multiple ancestries. Nat Genet. 2019;51:481–93.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  57. UK Biobank. Neale lab, http://www.nealelab.is/uk-biobank. Accessed 4 Dec 2023.

  58. Lenschow DJ, Lai C, Frias-Staheli N, et al. IFN-stimulated gene 15 functions as a critical antiviral molecule against influenza, herpes, and Sindbis viruses. Proc Natl Acad Sci U S A. 2007;104:1371–6.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  59. Shin D, Mukherjee R, Grewe D, et al. Papain-like protease regulates SARS-CoV-2 viral spread and innate immunity. Nature. 2020;587:657–62.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  60. Fujii W, Kapellos TS, Baßler K, et al. Alveolar macrophage transcriptomic profiling in COPD shows major lipid metabolism changes. ERJ Open Res. 2021;7:00915–2020.

    Article  PubMed  PubMed Central  Google Scholar 

  61. Bingle L, Wilson K, Musa M, et al. BPIFB1 (LPLUNC1) is upregulated in cystic fibrosis lung disease. Histochem Cell Biol. 2012;138:749–58.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  62. Ohlmeier S, Mazur W, Linja-Aho A, et al. Sputum proteomics identifies elevated PIGR levels in smokers and mild-to-moderate COPD. J Proteome Res. 2012;11:599–608.

    Article  CAS  PubMed  Google Scholar 

  63. Gao J, Ohlmeier S, Nieminen P, et al. Elevated sputum BPIFB1 levels in smokers with chronic obstructive pulmonary disease: a longitudinal study. Am J Physiol Lung Cell Mol Physiol. 2015;309:L17-26.

    Article  CAS  PubMed  Google Scholar 

  64. Yu W, Ye T, Ding J, et al. miR-4456/CCL3/CCR5 pathway in the pathogenesis of tight junction impairment in chronic obstructive pulmonary disease. Front Pharmacol. 2021;12:551839.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  65. Ravi AK, Khurana S, Lemon J, et al. Increased levels of soluble interleukin-6 receptor and CCL3 in COPD sputum. Respir Res. 2014;15:103.

    Article  PubMed  PubMed Central  Google Scholar 

  66. Drenth JP, Göertz J, Daha MR, et al. Immunoglobulin D enhances the release of tumor necrosis factor-alpha, and interleukin-1 beta as well as interleukin-1 receptor antagonist from human mononuclear cells. Immunology. 1996;88:355–62.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  67. Offord KP, Gleich GJ, Barbee RA, et al. Serum IgD in subjects with and without chronic obstructive pulmonary disease: a previous finding restudied. Am Rev Respir Dis. 1982;126:118–20.

    CAS  PubMed  Google Scholar 

  68. Shi R, Cao Z, Li H, et al. Peroxidasin contributes to lung host defense by direct binding and killing of gram-negative bacteria. PLoS Pathog. 2018;14:e1007026.

    Article  PubMed  PubMed Central  Google Scholar 

  69. Wu X, Sun X, Chen C, et al. Dynamic gene expressions of peripheral blood mononuclear cells in patients with acute exacerbation of chronic obstructive pulmonary disease: a preliminary study. Crit Care Lond Engl. 2014;18:508.

    Article  Google Scholar 

  70. Sueblinvong V, Liangpunsakul S. Relationship between serum leptin and chronic obstructive pulmonary disease in US adults: results from the third National Health and Nutrition Examination Survey. J Investig Med Off Publ Am Fed Clin Res. 2014;62:934–7.

    CAS  Google Scholar 

  71. Takabatake N, Nakamura H, Abe S, et al. Circulating leptin in patients with chronic obstructive pulmonary disease. Am J Respir Crit Care Med. 1999;159:1215–9.

    Article  CAS  PubMed  Google Scholar 

  72. Breyer M-K, Rutten EPA, Vernooy JHJ, et al. Gender differences in the adipose secretome system in chronic obstructive pulmonary disease (COPD): a pivotal role of leptin. Respir Med. 2011;105:1046–53.

    Article  PubMed  Google Scholar 

  73. Garcia IPL, Alfaro-Arnedo E, Canalejo M, et al. Insulin-Like Growth Factors and IGF-Binding Proteins levels in serum from smokers and patients with different grades of COPD, COPD and lung cancer and exacerbated COPD. Eur Respir J. 58. https://doi.org/10.1183/13993003.congress-2021.PA1952. Epub ahead of print 5 September 2021.

  74. Insuela DBR, Azevedo CT, Coutinho DS, et al. Glucagon reduces airway hyperreactivity, inflammation, and remodeling induced by ovalbumin. Sci Rep. 2019;9:6478.

    Article  PubMed  PubMed Central  Google Scholar 

  75. Zhou A, Carrell RW, Murphy MP, et al. A redox switch in angiotensinogen modulates angiotensin release. Nature. 2010;468:108–11.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  76. Uhal BD, Li X, Piasecki CC, et al. Angiotensin signalling in pulmonary fibrosis. Int J Biochem Cell Biol. 2012;44:465–8.

    Article  CAS  PubMed  Google Scholar 

  77. Wang R, Zagariya A, Ibarra-Sunga O, et al. Angiotensin II induces apoptosis in human and rat alveolar epithelial cells. Am J Physiol. 1999;276:L885-889.

    CAS  PubMed  Google Scholar 

  78. Mancini GBJ, Etminan M, Zhang B, et al. Reduction of morbidity and mortality by statins, angiotensin-converting enzyme inhibitors, and angiotensin receptor blockers in patients with chronic obstructive pulmonary disease. J Am Coll Cardiol. 2006;47:2554–60.

    Article  CAS  PubMed  Google Scholar 

  79. Mortensen EM, Copeland LA, Pugh MJV, et al. Impact of statins and ACE inhibitors on mortality after COPD exacerbations. Respir Res. 2009;10:45.

    Article  PubMed  PubMed Central  Google Scholar 

  80. Zhang H, Kho AT, Wu Q, et al. CRISPLD2 (LGL1) inhibits proinflammatory mediators in human fetal, adult, and COPD lung fibroblasts and epithelial cells. Physiol Rep. 2016;4:e12942.

    Article  PubMed  PubMed Central  Google Scholar 

  81. Pratte KA, Curtis JL, Kechris K, et al. Soluble receptor for advanced glycation end products (sRAGE) as a biomarker of COPD. Respir Res. 2021;22:127.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  82. Joo D-H, Lee K-H, Lee C-H, et al. Developmental endothelial locus-1 as a potential biomarker for the incidence of acute exacerbation in patients with chronic obstructive pulmonary disease. Respir Res. 2021;22:297.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  83. Oyewumi L, Kaplan F, Sweezey NB. Lgl1, a mesenchymal modulator of early lung branching morphogenesis, is a secreted glycoprotein imported by late gestation lung epithelial cells. Biochem J. 2003;376:61–9.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  84. Himes BE, Jiang X, Wagner P, et al. RNA-Seq transcriptome profiling identifies CRISPLD2 as a glucocorticoid responsive gene that modulates cytokine function in airway smooth muscle cells. PLoS One. 2014;9:e99625.

    Article  PubMed  PubMed Central  Google Scholar 

  85. Lan J, Ribeiro L, Mandeville I, et al. Inflammatory cytokines, goblet cell hyperplasia and altered lung mechanics in Lgl1+/- mice. Respir Res. 2009;10:83.

    Article  PubMed  PubMed Central  Google Scholar 

  86. Buchwald J, Chenoweth MJ, Palviainen T, et al. Genome-wide association meta-analysis of nicotine metabolism and cigarette consumption measures in smokers of European descent. Mol Psychiatry. 2021;26:2212–23.

    Article  CAS  PubMed  Google Scholar 

  87. Landini A, Trbojević-Akmačić I, Navarro P, et al. Genetic regulation of post-translational modification of two distinct proteins. Nat Commun. 2022;13:1586.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  88. Li X, Zhou G, Tian X, et al. The polymorphisms of FGFR2 and MGAT5 affect the susceptibility to COPD in the Chinese people. BMC Pulm Med. 2021;21:129.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  89. Sun BB, Maranville JC, Peters JE, et al. Genomic atlas of the human plasma proteome. Nat Epub ahead print. 2018. https://doi.org/10.1038/s41586-018-0175-2.

    Article  Google Scholar 

  90. Võsa U, Claringbould A, Westra H-J, et al. Large-scale cis- and trans-eQTL analyses identify thousands of genetic loci and polygenic scores that regulate blood gene expression. Nat Genet. 2021;53:1300–10.

    Article  PubMed  PubMed Central  Google Scholar 

  91. Knigshoff M, Kneidinger N, Eickelberg O. TGF-ß signaling in COPD: deciphering genetic and cellular susceptibilities for future therapeutic regimen. Swiss Med Wkly. https://doi.org/10.4414/smw.2009.12528. Epub ahead of print 3 Oct 2009.

  92. Zeng Z, Li M, Wang M, et al. Increased expression of Siglec-9 in chronic obstructive pulmonary disease. Sci Rep. 2017;7:10116.

    Article  PubMed  PubMed Central  Google Scholar 

  93. Ishii T, Angata T, Wan ES, et al. Influence of SIGLEC9 polymorphisms on COPD phenotypes including exacerbation frequency: SIGLEC9 polymorphisms and COPD. Respirology. 2017;22:684–90.

    Article  PubMed  Google Scholar 

  94. Hill AC, Guo C, Litkowski EM, et al. Large scale proteomic studies create novel privacy considerations. Sci Rep. 2023;13:9254.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  95. Singh D, Fox SM, Tal-Singer R, et al. Induced sputum genes associated with spirometric and radiological disease severity in COPD ex-smokers. Thorax. 2011;66:489–95.

    Article  PubMed  Google Scholar 

  96. Vucic EA, Chari R, Thu KL, et al. DNA methylation is globally disrupted and associated with expression changes in chronic obstructive pulmonary disease small airways. Am J Respir Cell Mol Biol. 2014;50:912–22.

    Article  PubMed  PubMed Central  Google Scholar 

  97. Herpel LB, Kanner RE, Lee SM, et al. Variability of spirometry in chronic obstructive pulmonary disease. Am J Respir Crit Care Med. 2006;173:1106–13.

    Article  PubMed  PubMed Central  Google Scholar 

  98. Magnussen H, Vaz Fragoso CA, Miller MR, et al. Spirometry variability must be critically interpreted before negating a clinical diagnosis of chronic obstructive pulmonary disease. Am J Respir Crit Care Med. 2018;197:835–6.

    Article  PubMed  Google Scholar 

  99. Emeny RT, Carpenter DO, Lawrence DA. Health disparities: Intracellular consequences of social determinants of health. Toxicol Appl Pharmacol. 2021;416:115444.

    Article  CAS  PubMed  Google Scholar 

  100. Bao X, Borné Y, Yin S, et al. The associations of self-rated health with cardiovascular risk proteins: a proteomics approach. Clin Proteom. 2019;16:40.

    Article  Google Scholar 

Download references

Acknowledgements

Molecular data for the Trans-Omics in Precision Medicine (TOPMed) program was supported by the National Heart, Lung and Blood Institute (NHLBI). Genome Sequence data for "NHLBI TOPMed: COPDGene" (phs000951) was performed at Broad Genomics and the Northwest Genome Center at the University of Washington (NWGC) (HHSN268201500014C, 3R01HL089856-08S1). Core support including centralized genomic read mapping and genotype calling, along with variant quality metrics and filtering were provided by the TOPMed Informatics Research Center (3R01HL-117626-02S1; contract HHSN268201800002I). Core support including phenotype harmonization, data management, sample-identity QC, and general program coordination were provided by the TOPMed Data Coordinating Center (R01HL-120393; U01HL-120393; contract HHSN268201800001I). We gratefully acknowledge the studies and participants who provided biological samples and data for TOPMed. The authors thank the SPIROMICS participants and participating physicians, investigators, study coordinators, and staff for making this research possible. More information about the study and how to access SPIROMICS data is available at www.spiromics.org. The authors would like to acknowledge the University of North Carolina at Chapel Hill BioSpecimen Processing Facility (http://bsp.web.unc.edu/) and Alexis Lab (https://www.med.unc.edu/cemalb/facultyresearch/alexislab/) for sample processing, storage, and sample disbursements. We would like to acknowledge the following current and former investigators of the SPIROMICS sites and reading centers: Neil E Alexis, MD; Wayne H Anderson, PhD; Mehrdad Arjomandi, MD; Igor Barjaktarevic, MD, PhD; R Graham Barr, MD, DrPH; Patricia Basta, PhD; Lori A Bateman, MS; Christina Bellinger, MD; Surya P Bhatt, MD; Eugene R Bleecker, MD; Richard C Boucher, MD; Russell P Bowler, MD, PhD; Russell G Buhr, MD, PhD; Stephanie A Christenson, MD; Alejandro P Comellas, MD; Christopher B Cooper, MD, PhD; David J Couper, PhD; Gerard J Criner, MD; Ronald G Crystal, MD; Jeffrey L Curtis, MD; Claire M Doerschuk, MD; Mark T Dransfield, MD; M Bradley Drummond, MD; Christine M Freeman, PhD; Craig Galban, PhD; Katherine Gershner, DO; MeiLan K Han, MD, MS; Nadia N Hansel, MD, MPH; Annette T Hastie, PhD; Eric A Hoffman, PhD; Yvonne J Huang, MD; Robert J Kaner, MD; Richard E Kanner, MD; Mehmet Kesimer, PhD; Eric C Kleerup, MD; Jerry A Krishnan, MD, PhD; Wassim W Labaki, MD; Lisa M LaVange, PhD; Stephen C Lazarus, MD; Fernando J Martinez, MD, MS; Merry-Lynn McDonald, PhD; Deborah A Meyers, PhD; Wendy C Moore, MD; John D Newell Jr, MD; Elizabeth C Oelsner, MD, MPH; Jill Ohar, MD; Wanda K O’Neal, PhD; Victor E Ortega, MD, PhD; Robert Paine, III, MD; Laura Paulin, MD, MHS; Stephen P Peters, MD, PhD; Cheryl Pirozzi, MD; Nirupama Putcha, MD, MHS; Sanjeev Raman, MBBS, MD; Stephen I Rennard, MD; Donald P Tashkin, MD; J Michael Wells, MD; Robert A Wise, MD; and Prescott G Woodruff, MD, MPH. The project officers from the Lung Division of the National Heart, Lung, and Blood Institute were Lisa Postow, PhD, and Lisa Viviano, BSN; SPIROMICS was supported by contracts from the NIH/NHLBI (HHSN268200900013C, HHSN268200900014C, HHSN268200900015C, HHSN268200900016C, HHSN268200900017C, HHSN268200900018C, HHSN268200900019C, HHSN268200900020C), grants from the NIH/NHLBI (U01 HL137880, U24 HL141762, R01 HL182622, and R01 HL144718), and supplemented by contributions made through the Foundation for the NIH and the COPD Foundation from Amgen; AstraZeneca/MedImmune; Bayer; Bellerophon Therapeutics; Boehringer-Ingelheim Pharmaceuticals, Inc.; Chiesi Farmaceutici S.p.A.; Forest Research Institute, Inc.; Genentech; GlaxoSmithKline; Grifols Therapeutics, Inc.; Ikaria, Inc.; MGC Diagnostics; Novartis Pharmaceuticals Corporation; Nycomed GmbH; Polarean; ProterixBio; Regeneron Pharmaceuticals, Inc.; Sanofi; Sunovion; Takeda Pharmaceutical Company; and Theravance Biopharma and Mylan/Viatris.

Funding

This work was supported by NHLBI R01 HL152735, U01 HL089897 and U01 HL089856. The COPDGene study (NCT00608764) is also supported by the COPD Foundation through contributions made to an Industry Advisory Committee that has included AstraZeneca, Bayer Pharmaceuticals, Boehringer-Ingelheim, Genentech, GlaxoSmithKline, Novartis, Pfizer, and Sunovion. COPDGene proteomics profiling was funded by through R01 HL137995 (Bowler, Kechris). SPIROMICS proteomic sample profiling was funded by Novartis.

Author information

Author notes

  1. Iain R Konigsberg and Thao Vu share first authorship

Authors and Affiliations

  1. Department of Biomedical Informatics, School of Medicine, University of Colorado – Anschutz Medical Campus, Aurora, CO, USA

    Iain R Konigsberg, Elizabeth M Litkowski, Luciana B Vargas, Niles Gilmore & Leslie A Lange

  2. Department of Biostatistics and Informatics, Colorado School of Public Health, Aurora, CO, USA

    Thao Vu, Weixuan Liu & Katerina J Kechris

  3. Department of Medicine, University of Michigan, Ann Arbor, MI, USA

    Elizabeth M Litkowski

  4. Department of Medicine, National Jewish Health, Denver, CO, USA

    Katherine A Pratte & Russell P Bowler

  5. Department of Computer Science and Engineering, University of Colorado – Denver, Denver, CO, USA

    Mohamed Abdel-Hafiz & Farnoush Banaei-Kashani

  6. Center for Public Health Genomics, University of Virginia, Charlottesville, VA, USA

    Ani Manichaikul

  7. Channing Division of Network Medicine, Division of Pulmonary and Critical Care Medicine, Brigham and Women’s Hospital, Harvard Medical School, Boston, MA, USA

    Michael H Cho, Craig P Hersh & Dawn L DeMeo

Authors
  1. Iain R Konigsberg
    View author publications

    You can also search for this author in PubMed Google Scholar

  2. Thao Vu
    View author publications

    You can also search for this author in PubMed Google Scholar

  3. Weixuan Liu
    View author publications

    You can also search for this author in PubMed Google Scholar

  4. Elizabeth M Litkowski
    View author publications

    You can also search for this author in PubMed Google Scholar

  5. Katherine A Pratte
    View author publications

    You can also search for this author in PubMed Google Scholar

  6. Luciana B Vargas
    View author publications

    You can also search for this author in PubMed Google Scholar

  7. Niles Gilmore
    View author publications

    You can also search for this author in PubMed Google Scholar

  8. Mohamed Abdel-Hafiz
    View author publications

    You can also search for this author in PubMed Google Scholar

  9. Ani Manichaikul
    View author publications

    You can also search for this author in PubMed Google Scholar

  10. Michael H Cho
    View author publications

    You can also search for this author in PubMed Google Scholar

  11. Craig P Hersh
    View author publications

    You can also search for this author in PubMed Google Scholar

  12. Dawn L DeMeo
    View author publications

    You can also search for this author in PubMed Google Scholar

  13. Farnoush Banaei-Kashani
    View author publications

    You can also search for this author in PubMed Google Scholar

  14. Russell P Bowler
    View author publications

    You can also search for this author in PubMed Google Scholar

  15. Leslie A Lange
    View author publications

    You can also search for this author in PubMed Google Scholar

  16. Katerina J Kechris
    View author publications

    You can also search for this author in PubMed Google Scholar

Contributions

IK, TV, WL, EL, KP, KK analyzed and interpreted proteomics data. IK, EL, LV, NG performed QTL analyses. IK, TV, WL, KP, LV, NG, KK contributed to the writing of the initial draft. IK, TV, WL, EL, KP, LV, NG, MH, AM, MC, CH, DD, FK, RB, LL, KK participated in the reviewing and editing process. All authors have read and approved the final manuscript.

Corresponding author

Correspondence to Katerina J Kechris.

Ethics declarations

Ethics approval and consent to participate

This study was conducted in accordance with the amended Declaration of Helsinki. The NIH-sponsored multicenter Genetic Epidemiology of COPD (COPDGene) study was approved and reviewed by the institutional review board (ClinicalTrials.gov Identifier: NCT00608764) including: National Jewish IRB, Partners Human Research Committee, Institutional Review Board for Baylor College of Medicine and Affiliated Hospitals, Institutional Review Board for Baylor College of Medicine and Affiliated Hospitals, Columbia University Medical Center IRB, The Duke University Health System Institutional Review Board for Clinical Investigations (DUHS IRB), Johns Hopkins Medicine Institutional Review Boards (JHM IRB), The John F. Wolf, MD Human Subjects Committee of Harbor UCLA Medical Center, Morehouse School of Medicine Institutional Review Board, Temple University Office for Human Subjects Protections Institutional Review Board, The University of Alabama at Birmingham Institutional Review Board for Human Use, University of California, San Diego Human Research Protections Program, The University of Iowa Human Subjects Office, VA Ann Arbor Healthcare System IRB, University of Minnesota Research Subjects’ Protection Programs (RSPP), University of Pittsburgh Institutional Review Board, UT Health Science Center San Antonio Institutional Review Board, Health Partners Research Foundation Institutional Review Board, Medical School Institutional Review Board (IRBMED), Minneapolis VAMC IRB Institutional Review Board/Research Review Committee Saint Vincent Hospital Fallon Clinic Fallon Community Health Plan.

All study participants provided written informed consent.

Consent for publication

Not applicable

Competing interests

The authors declare no competing interests.

Additional information

Publisher’s Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary Information

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Konigsberg, I.R., Vu, T., Liu, W. et al. Proteomic networks and related genetic variants associated with smoking and chronic obstructive pulmonary disease. BMC Genomics 25, 825 (2024). https://doi.org/10.1186/s12864-024-10619-1

Download citation

  • Received:

  • Accepted:

  • Published:

  • DOI: https://doi.org/10.1186/s12864-024-10619-1

Keywords

BMC Genomics

ISSN: 1471-2164

Contact us