Machine learning-based pan-cancer study of classification and mechanism of BRAF inhibitor resistance.

Background: V-raf murine sarcoma viral oncogene homolog B1 (BRAF) inhibitor (BRAFi) therapy resistance affects approximately 15% of cancer patients, leading to disease recurrence and poor prognosis. The aim of the study was to develop a machine-learning based method to identify patients who are at high-risk of BRAFi resistance and potential biomarker.

Methods: From Cancer Cell Line Encyclopedia (CCLE) and Genomics of Drug Sensitivity in Cancer (GDSC) databases, we collected RNA sequencing and half maximal inhibitory concentration (IC₅₀) data from 235 pan-cancer cell lines and then identified 37 significant differential expression genes associated with BRAFi resistance. Employing machine learning (ML) models, we successfully classified cell lines into resistant and sensitive groups, achieving robust performance in external validation datasets.

Results: AOX1 may play a vital part in BRAFi metabolism and resistance. Further, we found that higher mRNA expression of OXTR, H2AC13, and TBX2, and lower mRNA of SLC2A4, as detected by PCR in WM983B and SKMEL-5 cell lines, were independent risk factors for BRAFi resistance and were associated with poor prognosis.

Conclusions: We established a gene-expression model using ML methods, consisting of 37 variables based on RNA-seq database, which was externally validated and could be used to predict BRAFi resistance. Meanwhile, our findings provide valuable insights into the molecular mechanisms of BRAFi resistance, enabling the identification of high-risk patients.