A comprehensive meta QTL analysis for fiber quality, yield, yield related and morphological traits, drought tolerance, and disease resistance in tetraploid cotton

Background The study of quantitative trait loci (QTL) in cotton (Gossypium spp.) is focused on traits of agricultural significance. Previous studies have identified a plethora of QTL attributed to fiber quality, disease and pest resistance, branch number, seed quality and yield and yield related traits, drought tolerance, and morphological traits. However, results among these studies differed due to the use of different genetic populations, markers and marker densities, and testing environments. Since two previous meta-QTL analyses were performed on fiber traits, a number of papers on QTL mapping of fiber quality, yield traits, morphological traits, and disease resistance have been published. To obtain a better insight into the genome-wide distribution of QTL and to identify consistent QTL for marker assisted breeding in cotton, an updated comparative QTL analysis is needed. Results In this study, a total of 1,223 QTL from 42 different QTL studies in Gossypium were surveyed and mapped using Biomercator V3 based on the Gossypium consensus map from the Cotton Marker Database. A meta-analysis was first performed using manual inference and confirmed by Biomercator V3 to identify possible QTL clusters and hotspots. QTL clusters are composed of QTL of various traits which are concentrated in a specific region on a chromosome, whereas hotspots are composed of only one trait type. QTL were not evenly distributed along the cotton genome and were concentrated in specific regions on each chromosome. QTL hotspots for fiber quality traits were found in the same regions as the clusters, indicating that clusters may also form hotspots. Conclusions Putative QTL clusters were identified via meta-analysis and will be useful for breeding programs and future studies involving Gossypium QTL. The presence of QTL clusters and hotspots indicates consensus regions across cultivated tetraploid Gossypium species, environments, and populations which contain large numbers of QTL, and in some cases multiple QTL associated with the same trait termed a hotspot. This study combines two previous meta-analysis studies and adds all other currently available QTL studies, making it the most comprehensive meta-analysis study in cotton to date.


Background
The Gossypium genus is composed of approximately 50 species including four cultivated ones which vary in morphological and economic characteristics considerably [1]. Species which possess superior fiber and yield traits are tetraploids, Gossypium hirsutum and G. barbadense. G. hirsutum, also called Upland cotton, is grown in warm climates worldwide, and possesses high lint yield which is vital to the textile industry. G. barbadense, also called Egyptian cotton, Pima cotton, or Sea-island cotton, is known for superior fiber length, strength, and fineness [1]. Both species have been the focus of breeding programs to combine their superior fiber quantity and quality traits [1] since the rediscovery of Mendelian genetics more than a century ago.
In cotton and other crop species quantitative traits are governed by a multitude of loci each of which contributes to a particular phenotype in varying degrees [2]. These loci are called quantitative trait loci (QTL) and are evaluated for their positions on the chromosomes, and the phenotypic variance that they contribute to a particular trait [2]. Studies in Gossypium have focused on QTL which are involved in fiber strength, length, uniformity, micronaire, color, disease resistance, fruiting nodes, boll weight and number, yield, seed oil and protein content, leaf morphology, and various seed related traits [3,4]. Numerous independent studies have reported QTL pertaining to all of these traits using independent or updated linkage maps from different or the same segregating populations.
Historically QTL mapping studies were expensive undertakings that may have yielded little to no results. Creation of sufficient markers to saturate the genome was expensive with no guarantee that the markers generated would be anywhere close to the QTL of interest [5]. Marker development has improved significantly and has become much less expensive [6]. Mapping strategies have also progressed from simple interval mapping (SIM) to more complex composite interval mapping (CIM), multiple QTL mapping (MQM), and others which greatly increase the accuracy of placing QTL [7]. This study uses a mixture of more advanced mapping techniques including CIM and MQM.
QTL mapping is an important tool for breeders to combine economically important traits together to create a superior cultivar. Crosses between G. barbadense and G. hirsutum have been attempted for decades to combine their superior fiber quality and yield traits. While the two species are genetically close enough to be crossed there are issues with sterility, cytological abnormalities, and distorted segregation [4,8]. Differences in mapping populations, genotypes, and environments can yield heterogeneous results in QTL mapping [3,8]. For this reason, meta-analysis is useful in marker assisted selection as it merges datasets and creates consensus map positions for QTL. Not only does meta-analysis help to confirm the existence of declared QTL from various studies by creating possible hotspots where QTL from the same trait aggregate but it can imply the existence of pleotropic traits by creating QTL clusters for various traits.
Previous QTL mapping studies, 42 of which were cumulatively used in this study are summarized in Table 1 [4,. Fiber quality traits which are of upmost importance to cotton breeding programs have dominated these studies in Gossypium compared to yield and disease resistance related traits [3,4]. For this reason, the majority of the QTL in this study are fiber quality related QTL. Studies done thus far have varied in linkage groups used, mapping techniques, Gossypium species, mapping populations, and markers. It is important to consolidate all of the studies into a consensus map to identify possible QTL clusters and hotspots.
Two large scale meta-analyses of Gossypium were conducted by Rong et al. [3] and Lacape et al. [4]. The study by Lacape et al. focused on fiber quality traits, while Rong et al. [3] study was more comprehensive where the traits included fiber and seed, quality, drought tolerance to leaf morphology, and bacterial blight (Xcm) resistance. Lacape et al.'s [4] study on fiber quality was based on a G. barbadense x G. hirsutum (Gb x Gh hereafter) recombinant inbred line (RIL) population, whereas Rong et al.'s study pooled QTL from a multitude of segregating populations (F2, F2:3, or BC2) tested in a single environment with no replicates. The study by Rong et al. [3] pooled a variety of QTL trait types but lacked sufficient QTL from any one trait to declare hotspots. Since Rong et al.'s [3] work in 2007; many new yields and resistance related QTL have been described by numerous publications, not included in Lacape et al's [4] work. This study combines the work from the two meta-QTL studies and numerous other recent publications to perform a metaanalysis based on 42 different studies using a total of 1,223 QTL. Furthermore, some declared clusters in Rong et al.'s [3] study consisted of only two QTL, whereas clusters in this study required four or more QTL to declare a cluster. Placing more stringent requirements on what constitutes a cluster decreases the chances of declaring a false positive cluster. Also with the addition of newly discovered QTL declaring clusters with so few QTL would place a cluster on nearly every segment of every chromosome. In this study, the majority of clusters contains far more than four QTL and signifies a heavily populated region of QTL on the chromosome. With the addition of new QTL previously not described by meta-analysis, many hotspot clusters were identified in this study which were previously unknown. This study provides a comprehensive and updated meta-analysis of Gossypium.

Fiber length (FL)
A total of 151 FL QTL were reported over the genome with the exceptions of chromosomes c2 and 22 which contained none. Chromosomes c3, 23, and 26 contained the most QTL with 16, 10, and 14 QTL, respectively. Chromosomes c12 and 19 both contained 9 QTL, while chromosomes c1 and 14 both contained 8 QTL. Chromosome c4, 7, and 25 all contained 7 QTL, while chromosomes 6, 18, and 21 all contained 6 QTL. All other chromosomes contained 5 or less FL QTL.

Micronaire
Micronaire QTL were the most frequently identified with 234 QTL spread over the entire genome. Chromosomes Table 2 Distribution of fiber quality, yield, seed, leaf morphology and resistance QTL across the cotton genome (Continued)  The total QTL distributions on the genome are summarized below. The fiber quality, yield, seed, leaf morphology, and resistance related traits are summarized below.   QTL cluster and hotspot locations on the genome are described in the table below. Locations are given in cM regions which are seen graphically in Figure 1.

Fiber uniformity (FU)
A total of 91 FU QTL were distributed over the genome with the exception of chromosome c11 which contained none. The majority of QTL were found on chromosomes c12, 17, and 24 which contained 10, 7, and 6 QTL, respectively. All other chromosomes contained 5 or less FU QTL.

Fiber elongation (FE)
A total of 118 FE QTL was found across the genome excluding chromosome c11 which contained none. Chromosome c19 contained 12 QTL, while chromosomes c14, 15, 23, and 24 contained 8, 9, 8, and 9 QTL, respectively. Also notable were chromosomes c2 and 20, each of which contained 6 QTL. All other chromosomes contained 5 or less QTL.

Color
There were 71 total fiber color related QTL distributed over the genome with the exceptions of chromosomes c4, 12, 13, 20, and 24 which contained none. The most heavily populated chromosomes were chromosomes c6, 8, and 25 which contained 8, 9, and 11 QTL, respectively. Chromosome c9 contained 6 QTL, while all others contained 5 or less.

Short lint fibers (SLF)
A total of 2 SLF QTLs were identified and found on chromosomes c18 and 23 each.

Weight fitness (WT)
A total of 3 WT QTL were identified on chromosomes c10, 14, and 20.

Perimeter
A total of 2 perimeter QTL were identified on chromosomes c10 and 14.

Wall thickness
A total of 4 wall thickness QTL were identified to be distributed over 4 chromosomes (each on c3, 5, 9, and 18).

Lint index (LI)
A total of 15 LI QTL was found over the genome with no single chromosome having a notably higher quantity of QTL. Chromosomes c11, 12, 14, and 26 each contained 2 QTL, while chromosomes c4, 5, 7, 9, 10, 22, and 25 contained a single LI QTL.

Lint percent (LP)
A total of 25 QTL was found over the genome with no more than 3 QTL on any given chromosome.

Ratio of log (locule number) to log (boll number) (LB)
Two QTL was identified for this trait one on chromosome c23, and another on c25.

Harvest index (HI)
A total of 5 HI QTL were used in the study. Chromosome c2 and 18 contained 2 QTL each, while c14 contained only 1.

Seed Quality QTL Gossypol
A total of 6 gossypol QTLs were found over the genome with chromosome c19 having 2 QTL. Chromosomes c3, 13, 18, and 22 each contained only 1 gossypol QTL.

Hull percentage (HP)
A total of 2 HP QTL were used in the study and they were found on chromosomes c3 and 14.

Embryo protein percentage (EPP)
A total of 2 EPP QTL was used in the study and was found on chromosomes c6 and 15.

Number of large fiber seeds (LargenumFS)
Only 1 QTL was used for this trait and it was found on chromosome c12.

No fuzz fibers (NOFuzFib)
A total of 2 NoFuzFib QTL was identified both on chromosome 12.

Seed weight (SW)
A total of 5 SW QTL were used in the study. Chromosome c9 contained 2 QTL, while c7, 11, and 16 each contained 1 QTL.

Seed mass (SM)
Only 1 QTL was used for SM and was found on chromosome c2.

Pubescence
A total of 6 pubescence QTLs were identified. Chromosome c6 and 25 each contained 2 pubescence QTLs, while c1 and 23 only contained 1 QTL.
Height of node of first fruiting branch (HNFFB) Only 1 HNFFB QTL was used in the study and was found on chromosome c17.
Leaf morphology (size and shape) (LM) A total of 25 LM QTL was found across the genome which is associated with leaf size and shape.

Fusarium wilt resistance (Fusarium)
Only 2 Fusarium QTL were used in the study and they were found on chromosomes c2 and 15.

Xanthomonas campestris pv. Malvacearum (Xcm)
A total of 2 Xcm QTL was used in the study and were on chromosomes c5 and 14.

Physiology Related QTL Leaf chlorophyll
A total of 3 chlorophyll QTLs were used in the study, two of which were on chromosome c2, and one on c14.
Drought Tolerance QTL Trait Carbon isotope ratio (CIR) A total of 4 CIR QTL were used in the study. Chromosomes c14, 15, 17, and 25 each contained 1 CIR QTL.

Canopy temperature (CT)
Only 1 CT QTL was used in this study and was found on chromosome c6. The distribution of QTL on individual chromosomes can be analyzed from Table 2. A total of 1,223 QTL were distributed over the entire genome; however, some chromosomes contained high quantities of QTL compared to others. An even distribution of QTL would place about 47 QTL on each chromosome; however, this is not observed based on a Chi-square test (χ 2 =155.7 >37.65 at P = 0.05). Chromosomes c14 and 23 each contained 75 QTL, and were the most QTL densely populated chromosomes. They were followed by chromosomes c12, 24, and 25 which contained 61, 72, and 62 QTL, respectively. Also notable are chromosomes c5, 15, and 16 which carried 55, 53, and 57 QTL, respectively. Other chromosomes such as c4, 10, 11, and 20 carried relatively less QTL with 25, 27, 23, and 27 QTL, respectively.
Overall the A subgenome carried 536 QTL, while the D subgenome carried 687 QTL. The Chi-square test (χ 2 =18.72 >3.84 at P = 0.05) indicate that QTL were unevenly distributed on the two subgenomes with the D subgenome carrying significantly more QTL identified.
Chromosomes containing two or more QTL hotspots always contained at least 2 different trait QTL hotspot types. All hotspots overlapped with clusters in the same region. Some chromosomes containing more than one hotspot contained regions where hotspots overlapped forming hotspot clusters. For example, chromosome c4 contained an overlap between a FL and a micronaire QTL hotspot. Chromosome c6 contained an overlapping fiber color QTL hotspot and a micronaire QTL hotspot. Chromosome c7 contained an overlap between a FS, nematode resistance, and a FL QTL hotspot. Chromosome c12 contained an overlapping FU and micronaire QTL hotspots. Notably, chromosome c15 contained a FE, leaf morphology, and a micronaire QTL hotspot, all of which overlapped in the same region. Chromosome c16 contained an overlap between a VW resistance QTL hotspot and a micronaire QTL hotspot. Chromosome c23 contained two overlapping QTL hotspots for FE and VW resistance. Chromosome c24 contained an FS and a micronaire QTL hotspot which overlapped forming a fiber quality hotspot cluster. Chromosome c25 contained a fiber color and a micronaire QTL hotspot which overlapped forming another fiber quality hotspot cluster.

Discussion
The 1,223 QTL distributed over the Gossypium genome in this study revealed the presence of QTL clusters and specific trait QTL hotspots. Certain regions of the cotton chromosomes were more densely populated with QTL than other regions, implying the existence of QTL clusters. Hotspots for fiber quality, Verticillium resistance, nematode related resistance and leaf morphology were identified. Regions which contained clusters often contained hotspots, and in some cases hotspots overlapped. This data could potentially be used to identify new QTL in defined regions by clusters rather than attempting to scan the entire genome for QTL.
The significance of this study is the discovery of QTL clusters and hotspots, and reaffirmation of previous meta-analysis studies, which are of immediate value to breeders. The comprehensive cluster and hotspot map produced in this study will allow breeding programs to focus their efforts on regions of the genome containing the most QTL of interest. In addition, breeding programs may utilize the confidence interval data to only focus on QTL within narrow regions of the genome, as opposed to QTL containing broad confidence intervals. Hotspot clusters containing two or more hotspots will be of immediate interest to breeding programs as they contain multiple QTL of interest. In addition, this study is a much needed update of previous meta-analysis studies in Gossypium that is less comprehensive and up to date [3,4]. Previous meta-analyses have shown that QTL overlap to form clusters and that certain regions of the genome contain more QTL than others [3,4]. Therefore it is possible that novel QTL may be found within the clusters identified by this study.
In this study, data was collected from 42 different publications using a variety of population types (see Table 1), Gossypium species, environments, and QTL traits. Each study varied in their reporting of QTL in terms of confidence intervals used and LOD scores at which QTL were declared. The Biomercator V3 software requires confidence intervals (CI) in the QTL report; however, some publications did not include them. For those publications CI were estimated based on flanking marker positions of the QTL placed on the chromosome. Only QTL with current markers found in the cotton marker database and did not have multiple locations on a linkage groups were used in this study. When exact QTL positions were unknown and only flanking markers positions were given the average distance between markers was used to place the QTL on the chromosome. Also when exact confidence intervals were not provided, the flaking marker positions were used as confidence intervals. Each publication varied in terms of the average length of CI with some being significantly longer than others. The identification of clusters and hotspots are more problematic using QTL with wide CI. Not all QTL from every study was included in the analysis for the reasons described above to preserve accuracy in the study. Some studies identified QTL on A and D linkage groups and used flanking markers which did not coincide with markers known based on the Cotton Marker Database (CMD). For those QTL it was not possible to place them on the consensus map and they were therefore excluded from the study. QTL which were outside the range of the consensus map provided by the CMD were excluded from the study. Clusters were estimated within regions of approximately 20cM allowing for the possibility of QTL with wide confidence intervals to be possibly placed anywhere in that region. The Biomercator V3 software [50] also required LOD scores and R^2 (phenotypic variation explained by the QTL) values for all QTL in the study. Both the LOD scores and R^2 values varied wildly between studies; however, this data was irrelevant to the QTL placement on the chromosomes. QTL placement and inference of clusters were based only on the physical placement of the QTL in genetic distance (cM) given by each study, and the CI which was estimated by flanking markers if not given in the study.
The majority of the QTL data in this analysis was taken from fiber quality QTL analysis and not studies pertaining to yield traits or resistance QTL to biotic and abiotic stresses. Thus, the large QTL meta-analysis studies done in Gossypium have focused only on fiber quality traits [3,4]. This study pools together not only large scale fiber quality QTL studies but also yield, seed quality, leaf morphology, and resistance studies. The majority of hotspots were fiber quality related; however, five resistance, and one leaf morphology hotspot were identified. The existence of hotspots pertaining to multiple traits coexisting in the same region of a chromosome or hotspot clusters is beneficial to breeders, and some have been identified in this study. The impact of this study is the identification of fiber quality, yield, seed, leaf morphology, and resistance hotspots, and clusters containing various mixtures of all traits. The identification of clusters and hotspots will be useful in marker-assisted selection since the markers delineating these regions can be chosen for selecting the traits of interests in cotton breeding. For example, many of the QTL hotspots identified in this study contained hotspot cluster regions containing two or more hotspots pertaining to different traits. In this study, such regions often pertained to fiber quality related QTL hotspots. Breeding programs focused on fiber quality traits can focus on hotspot clustering regions and select for the flanking markers around the region. Most of the QTL clusters had hotspots associated with them, indicating that if new QTL are to be discovered they may be found around regions of known clusters. Marker assisted selection programs can utilize clusters found in this study to find novel QTL and possibly novel hotspots within the regions. This study identified regions of importance to marker assisted breeding programs with clusters and hotspots, and identified chromosomes which have the most QTL, clusters, and hotspots for future breeding programs.
To make the results publicly accessible, a database will be established. Updates will be also made regularly when new studies have been published. Users can access the information based on their needs or interests per trait or chromosome.

Conclusion
QTL clusters and hotspots were inferred and identified using the positions and distribution of QTL along the Gossypium genome. The presence of QTL clusters and hotspots indicate that genes pertaining to certain traits are more heavily concentrated in certain regions of the genome than others. Since multiple Gossypium species were used in the various publications it confirms that QTL clusters and hotspots are consistent throughout the genus. The study found clusters on every chromosome, but hotspots pertaining to specific traits are present only on some chromosomes. The locations of these clusters and hotspots will be beneficial for marker assisted selection and breeding programs focused on fiber quality, seed quality, disease resistance, leaf morphology, and other yield related traits. Fiber quality hotspots dominated this study, but five disease resistance hotspots, four pest resistance hotspots, and one leaf morphology hotspot were detected. The chromosomal locations of these clusters can be used as a starting point to identify new QTL using consensus marker and meta-analysis data. This may be useful for future QTL analysis to map regions of the genome with high phenotypic impact for various traits. To date this is the most comprehensive QTL meta-analysis study done with Gossypium as it utilizes past meta-analyses and current publications to identify novel clusters and hotspots previously not described.  Fiber color (Color): Fiber color is sometimes called reflectance or yellowness and is a measurement of the yellow tint in the fibers [52].

Methods
Fiber maturity (FM): Fiber maturity is described as the extent of development in the secondary walls [51]. Fiber perimeter (Perimeter): Fiber perimeter is a diverse trait which directly relates to fiber strength, length, and micronaire and is considered a fiber quality trait [53]. Weight fitness (WT): WT is a fiber quality trait relating to maturity which is measured by an arealometer, a device which measures compressional change in a known unit volume space with a known weighted sample [42]. Wall thickness (Wall Thick): Wall thickness is described as the thickness of the fiber in the secondary wall [51]. Short lint fibers (SLF): The content of short lint fibers is defined as percent by weight of fibers of 12.7 mm or less [52]. Harvest index (HI): The harvest index is a ratio by weight of seedcotton to dry matter produced by a cotton plant [53]. Boll weight (BW): Boll weight is the average weight in grams of a boll [51]. Lint index (LI): Lint index is described as the lint obtained from 100 seeds and is measured in grams [9]. LB is a calculation of the ratio of the log of locule number to the log of the number of bolls present [18]. Seed gossypol content (Gossypol): Gossypol is expressed as a percentage of gossypol in weight found in the cotton seed [54]. Seed protein content (Protein): Protein is expressed as the percentage of protein in weight found in the cotton seed [44]. Seed oil content (Oil): Oil is expressed as the percentage of oil in weight found in the cotton seed [44]. Hull percentage (HP): HP is an expression of the percentage of seed which comprises the hull of the embryo [9]. Embryonic protein percentage (EPP): EPP is a percentage of protein in weight present in the embryo [9]. Number of large fiber seeds (LargenumFS): Expressed as the number of large fiber seeds present. The trait relates to seed mass, with larger seeds having greater fitness than smaller seeds [5].
No fuzz fibers (NOFuzFib): No fuzz fibers is a trait described by seeds which only have lint fibers and no fuzz fibers [55]. Seed weight (SW): Seed weight is measured in grams [51]. Seed index (SI): Seed index is described as the weight of 100 seeds in grams [51]. Seed mass (SM): Seed mass is expressed as the seed mass in grams per unit area [3]. Number of fruiting branches (FB Num): Described as the number of fruiting branches per plant [14]. Number of fruiting branch nodes (FB Node): Described as the number of fruiting nodes per plant [14]. Node of first fruiting branch (NFFB): Used as a test of plant maturity and development by the examination of the maintem node (counted from the first true leaf ) of the first fruiting branch [20]. Height of node of first fruiting branch (HNFFB): A measurement of height of the first fruiting branch which relates to maturity and development traits [20]. Leaf morphology (Leaf Morph): Leaf morphology traits are described as differences between plants in leaf size, shape, and the number of lobes in each leaf [52]. Osmotic potential (OP): The osmotic potential is described as the plant's ability to adjust to osmotic differences via the active accumulation of solutes in response to a water deficit. This trait pertains to drought tolerance [27]. Nematode related resistance (Nematode Related): Nematode resistance related traits pertain to both reniform and root-knot nematodes. QTL for this trait are classified based on the number of eggs per gram root per plant, and the distribution of root galling index [35]. Fusarium wilt resistance (Fusarium): Fusarium resistance is a measurement of the plant's survival and resistance (disease severity rating) of infection by the inoculum [39]. Bacterial blight resistance (Xcm): Refers to traits which allow the plant to resist infection after being inoculated with Xcm [52]. Verticillium wilt (VW): Refers to traits (percentage of plants infected or disease severity rating) which allow the plant to resist infection after being inoculated with VW fungus [37]. Chlorophyll content (Chlorophyll): A measurement of the amount of chlorophyll present in the leaves. This trait describes QTL which appear to have a direct impact on chlorophyll content [27]. Carbon isotope ratio (CIR): A ratio of the carbon isotopes present which relates directly to a plant's ability to use water efficiently. CIR is a drought tolerance QTL trait [27].
Canopy temperature (CT): This trait is a measurement of canopy temperature and the amount of abiotic stress associated with the temperature [20]. CT is a drought tolerance related QTL trait [52].
Biomercator V3 [50] is capable of incorporating map files with QTL data files and displaying the results in a graphical representation. Map files consist of marker names along with the distance of the marker from the previous marker. The software then constructs the map based on adding the marker distances and displays the map along with markers with the appropriate distances between them. The QTL files consist of the map name, QTL name, chromosome number, trait, LOD score, phenotypic variance explained (R^2), mapping method used by the publication, position of the QTL, and the confidence interval.
The Biomercator V3 meta-analysis algorithm works by using a maximum likelihood method to calculate the most likely QTL distribution [56]. The CI, R^2, LOD scores, and positions of each QTL are assessed when calculating the existence of a cluster [56]. The algorithm assumes that each input QTL is not a false positive [56]. The algorithm then computes each possible model using the input QTLs and determines the most likely model [56]. Meta-analysis is a two-step process using Biomercator V3 [50]. First the linkage group on a specific chromosome is selected along with QTL of choice. During cluster analysis all QTL for a specific chromosome were included. The default kMax setting of 10 was used which in the second step allowed the software to calculate up to 10 possible clusters [50]. No chromosome in the study contained more than 5 clusters, so in the second step the program was instructed to find the best number of clusters appropriate for each chromosome based on the manual inference data.
Mapping methods varied between studies using CIM, MQM, and ICIM; however, this did not affect the Biomercator software's ability to place all QTL on the consensus map. Different QTL trait types are represented using different colors on chromosome maps. Each QTL is represented by a small horizontal line and a perpendicular vertical line. The horizontal line indicates the position in cM on the chromosome, and the vertical line represents the confidence interval of the QTL position on the map. The software is capable of calculating possible meta-clusters of QTL based on the number and position of QTL in a given region of the chromosome.
Meta-analysis was performed on each chromosome manually and using the software. Using manual inference both clusters and hotspots pertaining to specific traits were declared. These clusters and hotspots were then projected on the genome using Biomercator V3 software [50]. The same consensus map used for the Biomercator portion of the study was used and cluster and hotspot intervals were estimated based on marker positions on the map. The meta-analysis software used by Biomercator requires an input between 1 and 10 to display the "best meta-regions". For this reason the number of clusters estimated by the manual inference portion was used as input to the software. For example when the manual inference detected 2 clusters the software was configured to find the "best" 2 clusters. If the software was configured to find the "best" 3 or 4 clusters it often declared false positives, declaring a cluster when only 2 or 3 QTL were present. Hotspots were declared manually by removing all QTL trait types except for one to detect dense regions of that QTL type. Both clusters and hotspots were declared within approximately 20cM regions, meaning if a multiple QTL were detected between 0 and 20cM one cluster or hotspot was declared. This method of declaring clusters and hotspots with 20cM regions is based on the observation that large aggregates of QTL usually existed within a region about that size.