The integration of genomic information with clinical oncology presents a major challenge, due to the lack of functional understanding and statistical power of novel variants. Moreover, the rapid growth rate of cancer tumours may give rise to variants resistant to conventionally used therapies. To address these limitations, we have developed a suite of structural programs which predict the molecular consequences of coding variants on protein folding, stability and interactions. These can be used to provide insights into functional effects of such variants, and help guide patient clinical outcomes.
We have used this approach on gene variants in VHL and CDKN2B, where we accurately identified the patient risks of developing renal carcinomas based on protein stability and protein interaction effects. This was also possible for novel variants lacking clinical data. Following these predictions, 92% of the classified ‘high risk’ patients developed renal carcinoma, while none of the ‘low risk’ patients developed the disease. Applied to malignant paraganglioma, our structure-based variant analysis approach significantly predicted the chance of developing a malignant paraganglioma (p = 0.032) and patient life expectancy (p = 0.002).
When analysing over 200 first- and second- line kinase inhibitor resistant variants in ABL1 and EGFR, we found a significant reduction in binding affinity of the drugs relative to ATP binding. Higher frequency mutations were associated with milder effects on protein stability, which we relate to a lower fitness penalty. These structural insights enabled the development of a predictive model, which classified ABL1 and EGFR resistant variants with over 90% accuracy, as well as classifying resistance to either line of inhibitors. The mapping of these resistance patterns has revealed resistance hotspots, which aids in the design of resistance-resistant inhibitors.
These examples highlight how structural information, when applied in a systematic, integrated way, can provide a powerful and scalable approach to interpret genomic data to provide insights into clinical outcomes and patient management, and guiding future drug development.