Detection of microsatellite instability from circulating tumor DNA by targeted deep sequencing.

Abstract Currently, microsatellite instability (MSI) detection is limited to tissue samples with sufficient tumor content. Detection of MSI from the blood has been explored but confounded by low sensitivity due to limited circulating tumor DNA (ctDNA). In this study, we developed an NGS-based algorithm, bMSISEA (blood MSI signature enrichment analysis), to detect MSI from blood samples. BMSISEA development involved 3 major steps. First, marker sites that can effectively distinguish MSI-high (MSI-H) from microsatellite stable (MSS) tumors were extracted. Second, MSI signature enrichment analysis was performed based on hypergeometric probability, under the null hypothesis that the plasma sample has similar MSI-H and MSS read coverage pattern for a particular marker site as the white blood cells (WBC) from the training dataset. Finally, the enrichment score of each marker site was normalized and all markers were collectively considered to determine the MSI status of a plasma sample. In vitro dilution experiments with cell lines and in silico simulation experiments based on mixtures of MSI-H plasma and paired WBC DNA demonstrated 98% sensitivity and 100% specificity at a minimum of 1% ctDNA, and 91.8% sensitivity and 100% specificity with 0.4% ctDNA. An independent validation cohort of 87 colorectal cancer patients with orthogonal confirmation of MSI status of tissues confirmed the performance, achieving 94.1% sensitivity (16/17) and 100% specificity (27/27) for samples with ctDNA >0.4%.