Abstract
Human brain tissue studies have used a range of metrics to assess RNA quality but there are few large-scale cross-comparisons of presequencing quality metrics with RNA-seq quality. We analyzed how postmortem interval (PMI) and RNA integrity number (RIN) before RNA-seq relate to RNA quality after sequencing (percent of counts in top 10 genes [PTT], 5' bias, and 3' bias), and with individual gene counts across the transcriptome. We analyzed 4 human cerebrocortical tissue sets (1 surgical, 3 autopsy), sequenced with varying protocols. Postmortem interval and RIN had a low inverse correlation (down to r = -0.258, P < .001 across the autopsy cohorts); both PMI and RIN showed consistent and opposing correlations with PTT (up to r = 0.215, P < .001 for PMI and down to r = -0.677, P < .001 for RIN across the autopsy cohorts). Unlike PMI, RIN showed consistent correlations with measurements of 3' and 5' bias in autopsies (r = -0.366, P < .001 with 3' bias). RNA integrity number correlated with 3933 genes across the 4 datasets vs 138 genes for PMI. Neuronal and immune response genes correlated positively and negatively with RIN, respectively. Thus, different gene sets have divergent relationships with RIN. These analyses suggest that conventional metrics of RNA quality have varying values and that PMI has an overall modest effect on RNA quality.