Super-MSI: CfDNA-based pan-cancer microsatellite instability detection.

e13672 Background: Microsatellite instability (MSI) is a molecular subtype found in many cancers, which is often associated with the benefits of immunotherapy. However, the application of traditional testing strategies (such as PCR or IHC) is largely limited due to their dependencies on sufficient tumor tissues. Here, we introduced Super-MSI, a cfDNA-based pan-cancer MSI detection method using next-generation sequencing (NGS). Methods: 5500 detectable microsatellite loci from 250 white blood cells (WBCs) and 29 PCR-validated MSI-H tissue samples were sequenced and characterized, while the former represented for MSS pattern, and the later, for MSI-H. 91 microsatellite loci with the most variable motif repeat numbers between patterns were filtered and the Super-MSI method was established. Plasma samples were collected from 608 patients with known MSI status assessed by PCR (593 MSS and 15 MSI-H) to validate the method. All samples were sequenced with a 642-gene panel, and initial microsatellite reads count distributions were profiled by MSIsensor. Results: We characterized each microsatellite locus by REF alleles (MSI-H:WBC ratio less than 1) and ALT alleles (MSI-H:WBC ratio more than 20), then defined the sum of ALT alleles frequency in WBC group as the baseline for each candidate locus. For plasma samples with unknown MSI status, hypergeometric distribution and FDR calibration were employed to assess the difference of ALT alleles frequency between the WBC baseline and plasma samples. Locus with FDR adjusted p-value less than 0.05 was defined as an unstable locus based on prior knowledge, and negative log-transformed FDR-p was defined as MSscore for the locus. The plasma sample with both unstable microsatellite proportion larger than 20% and total MSscore larger than 270 was defined as MSI-H, and contrariwise, MSS. For all evaluable samples in the validation cohort, Super-MSI showed high sensitivity and specificity of 60% (9/15) and 100% (593/593) for an overall accuracy of 99.0% (602/608), superior to its kind bMSISEA (83.5%, 106/127) and Guardant360 (98.4%, 934/949). Within LOD of 0.5% maxAF (approximately 1% ctDNA fraction), Super-MSI was able to detect 81.8% (9/11) of MSI-H samples, demonstrating high concordance with tissue-based PCR tests. Conclusions: Super-MSI can genotype pan-cancer patients with blood cfDNA samples and give highly accurate MSI evaluation with 81.8% sensitivity and 100% specificity above 1% ctDNA fraction.