Abstract 5349: Accurate molecular profiling of sequence and copy number alterations from sub-nanogram FFPE DNA amounts

Introduction: Formalin-fixed paraffin-embedded (FFPE) specimens represent an invaluable source of material for precision oncology. However, FFPE samples pose significant challenges for molecular assays, such as targeted Next-Generation Sequencing (NGS), because of the highly variable DNA quality and often limited sample size. Here we present a complete workflow from sample quality control to targeted library preparation to reliably detect sequence and copy number alterations (CNA) through targeted NGS from extremely low input FFPE samples. Materials & methods: Three FFPE specimens from patients with breast (BrCa) or pancreatic cancer, with DNA quality varying over a broad range and with low tumor cellularity (down to 10%) were selected for targeted NGS profiling. The DNA quality was determined using the DEPArray™ FFPE QC Kit, a qPCR-based assay yielding a QC score defined as the ratio between the quantification of a 132 bp amplicon, corresponding approximately to the average length (116bp) of DEPArray OncoSeek amplicons, and a shorter amplicon of 54 bp. The QC scores of the 3 FFPE specimens ranged between 0.23 to 0.74. We then used QC score, multiplied by the ploidy (assessed on DEPArray during cell sorting) and by the number of cells recovered, to estimate the effectively amplifiable template (EAT). Thirty two pure cell populations (21 stromal and 11 tumor), with different EATs (80-300), corresponding to a wide range of number of cells per pool (31-214), were collected with the DEPArray digital sorter. Libraries were prepared from each cell pool, using the single-tube, Illumina-compatible DEPArray OncoSeek panel comprising 63 oncology relevant genes (average 740,000 sequenced amplicons per sample). Results: Variant calling showed sensitivity comprised between 93% and 99% and specificity > 99% for EATs ≥ 80 equivalent to as low as ≈250 pg of DNA. Analysis of CNAs in stromal cells was highly specific (zero false positive at 1.5 fold-change threshold). Moreover, analysis allowed to highly reproducibly identify CNAs in CCND1 (3x), ERBB2 (8x), MYC (3x) and PIK3KA (2x) in both replicates of one BrCa sample. Similarly, ERBB2 amplification (7x) was found in the other BrCa sample (10% tumor cellularity) across two replicates at different EATs (80, 120) corresponding to as low as 59 and 75 cells. As expected by the low tumor content, in the corresponding unsorted sample ERBB2 was below the 2-fold gain threshold with respect to the stromal control, which would not q