Analyzing Protein Stability and Phosphorylations with Machine Learning – UROP Symposium

Analyzing Protein Stability and Phosphorylations with Machine Learning

Riya Jwalanna

Research Mentor: Sriram Chandrasekaran
Mentor Department: Biomedical Engineering, Medicine
Author(s): Riya Jwalanna, Jaie Woodard, Sriram Chandrasekaran
Session: Session 3 (11:00 AM – 11:50 AM)
Presentation Type: Poster 57

Abstract

Stability altering protein phosphorylations can skew protein activity and/or abundance leading to disease. One way to obtain data for studying the relationship between protein stability (G) and phosphorylation is to directly model phosphorylation using computational methods. However, this proves less accurate than employing phosphomimetic mutations – amino acid substitutions that mimic the negative charge of protein phosphorylation. Phosphomimetic and phosphorylation data was used to obtain predicted G values and a machine learning approach was selected to be trained on the data. We also incorporated tumor mutational burden (TMB) – an important marker for how tumors respond to immunotherapies. A higher TMB may indicate higher success with immunotherapy and conversely, a lower TMB may indicate lower success. Our findings showed that a lower tumor mutational burden was associated with a higher destability with a statistically significant p-value of 1E-7 at =0.05. The variance for ‘low TMB’ was higher than for ‘high TMB’ where the difference in variances was 0.0823. This would mean a lower amount of genetic mutations in cancer cells would be associated with higher rates of protein destability. This could be due to passenger phosphorylations, ‘passengers’ that exist but do not hinder or help cancer growth, accumulating more within the ‘high TMB’ category. Jia et al. found that “as the number of mutations increase, relatively fewer have functional impact.” If we apply this to our findings with protein phosphorylations in combination with mutations, to provide a more complete picture rather than only genetic mutations, we can conclude it is likely that the high TMB cancers may include many lower-impact passenger phosphorylations, whereas the cases with low TMB may include higher-impact phosphorylations that propagate the cancer.

lsa logoum logo