Analyzing allele specific RNA expression using mixture models

BMC Genomics

Table 1 Poisson mixture model parameter estimates and SNPs classification results

Mixture component	Proportion	Poisson mean	No. of SNPs	No. of genes
Comp.1	0.030 (0.029, 0.031)	43.11 (42.54, 43.84)	18367	784
Comp.2	0.0011 (0.0010, 0.0012)	152.37 (146.08, 166.13)	519	37
Comp.3	0.186 (0.182, 0.190)	20.34 (20.20, 20.49)	82963	3892
Comp.4	0.003 (0.0025, 0.0033)	108.14 (105.13, 115.60)	2073	89
Comp.5	0.0006 (0.0004, 0.0008)	201.01 (196.15, 209.71)	425	27
Comp.6	0.0073 (0.0069, 0.0077)	74.60 (72.56, 78.08)	5156	202
Comp.7	0.771 (0.769, 0.775)	7.82 (7.78, 7.85)	198889	11174

The Poisson mixture model was fitted to the averaged total reads within tissue-specific genes (62326 tissue-specific genes in total, i.e. sample size = 62326; overall log-likelihood = -216846; BIC = 433836). Genes with the same rs number but from different brain region were considered as different tissue-specific genes. We found the optimal number of mixture components to be 7, meaning that we could classify all SNPs into 7 “comparable” SNP groups. Most SNPs in the gene of our interest (SLC1A3) were classified into the mixture component Comp.1. The SNPs in Comp.1 were used to fit the folded Skellam mixture model

ISSN: 1471-2164