Abstract
Transcriptome profiling can provide information of great value in clinical decision-making, yet RNA from readily available formalin-fixed paraffin-embedded (FFPE) tissue is often too degraded for quality sequencing. To assess the clinical utility of FFPE-derived RNA, we performed ribo-deplete RNA extractions on > 3200 FFPE slide samples; 25 of these had direct FFPE vs. fresh frozen (FF) replicates, 57 were sequenced in 2 different labs, 87 underwent multiple library analyses, and 16 had direct microdissected vs. macrodissected replicates. Poly-A versus ribo-depletion RNA extraction methods were compared using transcriptomes of TCGA cohort and 3116 FFPE samples. Compared to FF, FFPE transcripts coding for nuclear/cytoplasmic proteins involved in DNA packaging, replication, and protein synthesis were detected at lower rates and zinc finger family transcripts were of poorer quality. The greatest difference in extraction methods was in histone transcripts which typically lack poly-A tails. Encouragingly, the overall sequencing success rate was 81%. Exome coverage was highly concordant in direct FFPE and FF replicates, with 98% agreement in coding exon coverage and a median correlation of whole transcriptome profiles of 0.95. We provide strong rationale for clinical use of FFPE-derived RNA based on the robustness, reproducibility, and consistency of whole transcriptome profiling.
Highlights
Background correlationsformalin-fixed paraffin-embedded (FFPE) vs. The Cancer Genome Atlas (TCGA) FFPE vs.Extraction replicates High Transcript Integrity Number (TIN) (n data points = 57)Micro- vs. macrodissected replicates (n data points = 16)FFPE-fresh frozen (FF) replicatesHigh TIN (n data points = 25)FFPE samples from matched cancer type (n data points = 99,884)FFPE vs. TCGA after mapping cancer type matched (n data points = 631,353)FFPE vs. TCGA after mapping cancer type unmatched (n data points = 10,195,500)FFPE vs. TCGA before mapping cancer type matchedFFPE vs. TCGA before mapping cancer type unmatchedPossible scenarios for read distribution along a transcript
In examining a large cohort of FFPE patient samples for RNA sequencing we found that 86% of extracted RNA was successfully prepared for library prep and 94% of those samples had a sufficient amount of non-ribosomal RNA, resulting in an overall RNA sequencing success rate of 81%
As a part of the FFPE versus FF TCGA comparison, we developed a computational mapping methodology to project TCGA gene expression data into the same space as clinical FFPE cohort in order to correct for batch effects introduced by combining two such different datasets (Fig. 5Bi–ii)
Summary
FFPE vs TCGA FFPE vs. Extraction replicates High TIN (n data points = 57). Micro- vs macrodissected replicates (n data points = 16). FFPE-FF replicatesHigh TIN (n data points = 25). FFPE samples from matched cancer type (n data points = 99,884). FFPE vs TCGA after mapping cancer type matched (n data points = 631,353). FFPE vs TCGA after mapping cancer type unmatched (n data points = 10,195,500). FFPE vs TCGA before mapping cancer type matched. FFPE vs TCGA before mapping cancer type unmatched. Possible scenarios for read distribution along a transcript
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
