R-Score: A New Parameter to Assess the Quality of Variants' Calls Assessed by NGS Using Liquid Biopsies

• Published in: Biology (Basel, Switzerland) Vol. 10; no. 10; p. 954
• Main Authors: Serna-Blasco, Roberto, Sánchez-Herrero, Estela, Berrocal Renedo, María, Calabuig-Fariñas, Silvia, Molina-Vila, Miguel Ángel, Provencio, Mariano, Romero, Atocha
• Format: Journal Article
• Language: English
• Published: Switzerland MDPI AG 24.09.2021; MDPI
• Subjects: Alleles; Biopsy; ctDNA; DNA fingerprinting; DNA sequencing; Epidermal growth factor receptors; filtering; Genomes; Laboratories; liquid biopsy; Lung cancer; Mutation; Next-generation sequencing; NGS; Non-small cell lung carcinoma; Plasma; Polymerase chain reaction; Small cell lung carcinoma; Software; Tumors; VAF; variant calling; Palo Alto California; United States > US; VAF; variant calling; NGS; filtering; ctDNA; liquid biopsy
• ISSN: 2079-7737; 2079-7737
• DOI: 10.3390/biology10100954

Next-generation sequencing (NGS) has enabled a deeper knowledge of the molecular landscape in non-small cell lung cancer (NSCLC), identifying a growing number of targetable molecular alterations in key genes. However, NGS profiling of liquid biopsies risk for false positive and false negative calls and parameters assessing the quality of NGS calls remains lacking. In this study, we have evaluated the positive percent agreement (PPA) between NGS and digital PCR calls when assessing mutation status using 85 plasma samples from 82 -positive NSCLC patients. According to our data, variant allele fraction (VAF) was significantly lower in discordant calls and the median of the absolute values of all pairwise differences (MAPD) was significantly higher in discordant calls ( < 0.001 in both cases). Based on these results, we propose a new parameter that integrates both variables, named R-score. Next, we sought to evaluate the PPA for mutation calls between two independent NGS platforms using a subset of 40 samples from the same cohort. Remarkably, there was a significant linear correlation between the PPA and the R-score (r = 0.97; < 0.001). Specifically, the PPA of samples with an R-score ≤ -1.25 was 95.83%, whereas PPA falls to 81.63% in samples with R-score ≤ 0.25. In conclusion, R-score significantly correlates with PPA and can assist laboratory medicine specialists and data scientists to select reliable variants detected by NGS.