2022 World Conference on Lung Cancer (ePosters)-EP08.02-062. Evaluation of In Silico Tools to Determine Potential Actionability of Missense Variants with Experimental Therapies for NSCLC

EP08.02-062. Evaluation of In Silico Tools to Determine Potential Actionability of Missense Variants with Experimental Therapies for NSCLC

Back to course

Pdf Summary

Researchers from Northwestern University evaluated six widely used in silico tools for predicting the pathogenicity of missense variants in non-small cell lung cancer (NSCLC). The study focused on 15 genes that have preclinical or clinical evidence of actionability but lack FDA-approved targeted therapies. A total of 101 variants were analyzed using tools such as Polyphen-2, Align-GVGD, MutationTaster2021, CADD, CONDEL, and REVEL. The pathogenicity of each variant was determined using multiple databases, and variants with at least two concordant classifications were prioritized. The performance of the tools was evaluated based on overall accuracy, sensitivity, specificity, Matthews correlation coefficient (MCC), and likelihood ratio (LR). <br /><br />The results showed that all the tools had high sensitivity but low specificity. The overall accuracy ranged from moderate to high, and the MCCs were generally low. The LR values indicated minimal increase in the likelihood of disease for positive test results, while the LR- values showed a decrease in the likelihood of disease for negative test results. MutationTaster2021 demonstrated the highest level of performance with the highest specificity, MCC, LR, and lowest LR- scores. On the other hand, CONDEL had the lowest level of performance with the lowest accuracy, sensitivity, LR, and highest LR- scores.<br /><br />It was concluded that each in silico tool has different performance characteristics and limitations. While tools with high sensitivity can be useful in ruling out pathogenic variants, none of the tools had high enough specificity to effectively identify pathogenic variants. MutationTaster2021 was deemed the most reliable tool, but clinicians were advised to exercise caution in using these tools alone to determine actionability. The study provides valuable insights into the variability and limitations of in silico tools and emphasizes the need for integrating multiple sources of information in variant classification for precision medicine in NSCLC.

Asset Subtitle

Lena Chae

Meta Tag

Speaker Lena Chae

Topic Metastatic Non-small Cell Lung Cancer - Molecular Targeted Treatments

Keywords

in silico tools

pathogenicity prediction

missense variants

non-small cell lung cancer

gene actionability

variant analysis

performance evaluation

MutationTaster2021

CONDEL

precision medicine