CommonMind Consortium provides transcriptomic and epigenomic data for Schizophrenia and Bipolar Disorder

Abstract

Schizophrenia and bipolar disorder are serious mental illnesses that affect more than 2% of adults. While large-scale genetics studies have identified genomic regions associated with disease risk, less is known about the molecular mechanisms by which risk alleles with small effects lead to schizophrenia and bipolar disorder. In order to fill this gap between genetics and disease phenotype, we have undertaken a multi-cohort genomics study of postmortem brains from controls, individuals with schizophrenia and bipolar disorder. Here we present a public resource of functional genomic data from the dorsolateral prefrontal cortex (DLPFC; Brodmann areas 9 and 46) of 986 individuals from 4 separate brain banks, including 353 diagnosed with schizophrenia and 120 with bipolar disorder. The genomic data include RNA-seq and SNP genotypes on 980 individuals, and ATAC-seq on 269 individuals, of which 264 are a subset of individuals with RNA-seq. We have performed extensive preprocessing and quality control on these data so that the research community can take advantage of this public resource available on the Synapse platform at http://CommonMind.org .

Fig. 1

RNA-seq quality control metrics stratified by disease status. All samples from the 4 brain banks are shown.

Fig. 2

Integrated quality control of RNA-seq data. (a) Principal components analysis of log2 CPM values from RNA-seq across 4 brain banks. Brain bank, age of death, and diagnosis are indicated in the legend. (b) Plot of log2 CPM expression of UTY gene from chrY against XIST gene from chrX in order to validate reported sex.

Fig. 3

Sex check of ATAC-seq samples. (a) Heterozygosity rate of chromosome X genotype calls outside pseudoautosomal regions. (b) Read counts in OCRs on chromosome Y outside the pseudoautosomal region. (c,d) The read counts of OCRs adjacent to XIST and FIRRE genes.

Fig. 4

Quality control metrics for ATAC-seq samples. Histograms of (a) fraction of uniquely mapped reads (mean 0.919, sd ±0.010), (b) fraction of mitochondrial chromosome reads (mean 0.089, sd ±0.022), (c) mean insert sizes of pair-end reads (mean 288, sd ±29), (d) mean GC content (mean 0.418, sd ±0.012), (e) number of called peaks (mean 14,810, sd ±8,979), (f) mean coverage (mean 2.473, sd ±0.642), (g) normalized strand cross-correlation coefficient (mean 1.054, sd ±0.020), (h) relative strand correlation coefficient (mean 0.991, sd ±0.084)) and (i) fraction of fragments in peaks (FRiP) (mean 0.151, sd ±0.027).

Fig. 5

Summary of ATAC-seq data. (a) Genomic annotation of consensus OCRs (OCRs within 3 kb of a transcription start site were considered as promoter OCRs). (b) Clustering of the individual samples (n = 269) by chromatin accessibility in consensus OCRs using multidimensional scaling.

Fig. 6

Quality control of genotype data. Genotype QC for sex (a,b) and ancestry inference (c,d) for MSSM-Penn-Pitt (a,c) and HBCC (b,d). (a,b) F statistic from plink’s check-sex function, plotted by reported sex. Following data QC there is 100% concordance between reported sex and inferred sex based on F statistic for both MSSM-Penn-Pitt (a) and HBCC (b). (c,d) The first two principal components (PC) of genetic ancestry as inferred by GEMTOOLs. For both MSSM-Penn-Pitt (c) and HBCC (d) we see good concordance between reported ethnicity and genetic background clusters inferred by GEMTOOLs.

Fig. 7

Assessing sample concordance using genetic variants. Estimating contamination using Chipmix (x-axis) and Freemix (y-axis) output from VerifyBamID, on RNA-seq and genotyping data for HBCC and MSSM-Penn-Pitt cohorts. Each point is an RNA-seq sample and is colored according to whether the sample was accepted, excluded or rescued. Box in lower left-hand corner indicates criteria for a sample to be accepted if samples match the expected individual. Box in lower right indicates samples that were rescued by re-labeling to the proper individual. We note that this figure included samples there were excluded because of other filters.