Copy number variation (CNV) calling of MET, MYC, and ERBB2 using PiVAT bioinformatic platform.

Abstract

e15145 Background: CNV in tumor-associated genes such as MET, MYC and ERBB2 amplification can lead to cancer progression and poor survival in multiple cancers, and is associated with treatment regimen decision. Whole genome sequencing (WGS) and comprehensive genomic profiling (CGP) panels covering hundreds of genes are currently used to detect CNV in hospitals and CLIA-accredited laboratories. However, cost and complex workflow of WGS/CGP restrict their application in clinical molecular diagnostics. CGP panel covering dozens of genes based on targeted amplicon sequencing (TAS) method is preferred, but it is not ideal to detect CNV accurately due to unique methodology. Here we established a CNV calling toolkit in PiVAT bioinformatic platform to solve CNV calling problems in small-size TAS panel. Methods: 46 CNV positive (CNV+) and 14 CNV negative (CNV-) samples were sequenced using MGISEQ-2000 with a paired end, 2x100 read length protocol and adapter sequences were trimmed from 3’ ends of each read. PiVAT’s CNV calling method is based on double coverage normalization, including one per-sample normalization and one per-amplicon normalization. CNV- samples were used as normalization references. MET, MYC and ERBB2 copy number was investigated to determine sensitivity, specificity, precision, and accuracy of the overall test. Copy number deletion is defined as ratio < 0.8 and amplification is defined as ratio > 1.2. PiVAT CNV output calls were compared against excepted copy number to calculate and plot the concordance. CNVkit and Control-FREEC, 2 other CNV calling tools, were chosen as comparators to compare ERBB2, MET and MYC copy number levels with PiVAT’s CNV caller. Results: Sequencing results from all 46 CNV+ samples showed high coverage uniformity, with > 90% of sites having base coverage depth > 20% of mean coverage for all samples. Average mapping rate and on-target rate of sequencing run, analyzed by PiVAT, were 98.98±0.41% and 98.39±0.45%. Results from PiVAT showed a sensitivity of 97.83% and a specificity of 100% for MET, MYC, ERBB2 amplifications. PiVAT estimated copy number more accurately compared with CNVkit and Control-FREEC. PiVAT’s R2, measuring degree of difference from expected copy number, was greater than CNVkit and Control-FREEC for MET, MYC and ERBB2, with values at 0.78, 0.69 and 0.96, respectively. In contrast, CNVkit and Control-FREEC were on average 77.38% and 164.04% worse. For CNVkit and Control-FREEC, many of MET, MYC and ERBB2 CNV calls were seen at a much lower magnitude than expected. Conclusions: Our studies have shown that PiVAT is a reliable tool for calling CNVs. PiVAT’s sensitivity and accuracy were calculated to be > 97%. And PiVAT outperformed other CNV calling tools, CNVkit and Control-FREEC. This demonstrates that PiVAT’s CNV caller can be used to provide CNV results from TAS panel, and guide relevant cancer diagnosis and clinical decision more accurately.