Evaluating predictors of kinase activity of STK11 variants identified in primary human non-small cell lung cancers

Yile Chen ¹, Kyoungyeul Lee ², Junwoo Woo ²

¹Department of Biomedical Informatics and Medical Education, University of Washington, Seattle, 98105, WA, USA.
²3billion, 3billion Biotechnology company, Seoul, South Korea.
³Department of Pharmacy and Biotechnology, University of Bologna, Bologna, 40126, Italy.
⁴Department of Electrical and Computer Engineering, Texas A&M University, College Station, 77843, TX, USA.
⁵Molecular and Human Genetics, Baylor College of Medicine, Houston, 77030, TX, USA.
⁶Institute for Genomic Health, Icahn School of Medicine at Mount Sinai, New York, 10029, NY, USA.
⁷Department of Genetics and Genomic Sciences, Icahn School of Medicine at Mount Sinai, New York, 10029, NY, USA.
⁸Department of Pathology, University of Vermont, Burlington, 5445, VT, USA.
⁹Khoury College of Computer Sciences, Northeastern University, Boston, 02115, MA, USA.
¹⁰University of California, Berkeley, Berkeley, 94720, CA, USA.
¹¹Division of Genetics and Genomics, Boston Children's Hospital, Harvard Medical School, Boston, 02115, MA, USA.
¹²Broad Center for Mendelian Genomics, Program in Medical and Population Genetics, Broad Institute of Massachusetts Institute of Technology and Harvard, Cambridge, 02142, MA, USA.
¹³Center for Information Technology, National Institutes of Health, Bethesda, 20892, MD, USA.
¹⁴The Institute for Experiential AI, Northeastern University, Boston, 02115, MA, USA.

⁰Department of Biomedical Informatics and Medical Education, University of Washington, Seattle, 98105, WA, USA.

Research Square +

|

July 16, 2024

View abstract on PubMed

Summary

Computational tools accurately predict STK11 gene variant effects in lung cancer. The Critical Assessment of Genome Interpretation challenge showed these predictors aid in disease diagnosis and molecular discovery.

Area Of Science

Genomics
Computational Biology
Bioinformatics

Background

Computational tools are increasingly used for variant effect prediction in disease diagnosis and molecular discovery.
Accurate evaluation of these tools is crucial for reliable genomic interpretation.

Purpose Of The Study

To critically evaluate computational variant effect predictors using experimentally assayed STK11 rare variants from non-small cell lung cancer.
To assess the performance of participant and publicly available tools in the sixth Critical Assessment of Genome Interpretation (CAGI) challenge.

Main Methods

A dataset of 28 STK11 rare variants (27 missense, 1 deletion) from lung cancer biopsies was experimentally assayed.
Four participating teams and five public tools were evaluated on key metrics like correlation with assay outputs and separation of loss-of-function (LoF) from wildtype-like (WT-like) variants.
Functional data included biological and technical replicates to establish realistic maximum predictive performance.

Main Results

Predictors showed high performance in correlation and variant classification.
The best participant model, 3Cnet, was competitive with established tools.
Three public tools and 3Cnet approached assay replicate performance in distinguishing LoF from WT-like variants.
REVEL showed a correlation comparable to experimental replicates for real-valued assay output.

Conclusions

Combining functional, computational, and population data enabled 16 new variants to be classified as likely pathogenic or likely benign.
The STK11 challenge underscores the utility of variant effect predictors in biomedical research.
Results encourage further development in computational genome interpretation.

Keywords:

CAGI Peutz-Jeghers syndrome STK11 cancer kinase machine learning variant effect prediction

Related Concept Videos

PI3K/mTOR/AKT Signaling Pathway

The mammalian target of rapamycin (mTOR) is a serine/threonine kinase that regulates growth, proliferation, and cell survival in response to hormones, growth factors, or nutrient availability. This kinase exists in two structurally and functionally distinct forms: mTOR complex 1 (mTORC1) and mTOR complex 2 (mTORC2). The first form (mTORC1) is composed of a rapamycin-sensitive Raptor and proline-rich Akt substrate, PRAS40. In contrast, mTORC2 consists of a...