Immune checkpoint inhibitors: Interaction between PD-1 (blue) on a T-cell and PD-L1  (red) on a cancer cell  blocked by therapeutic antibodies
Credit: selvanegra/Getty Images

New research leveraging machine learning has identified novel predictive biomarkers of response to immune checkpoint inhibitor (ICI) therapy for melanoma, according to new research from the Wistar Institute published in Nature Communications. The investigators demonstrated that mutations in the processes of leukocyte and T-cell proliferation regulation show potential as biomarkers with reliable and stable prediction of ICI therapy response across multiple different datasets of melanoma patients.

“ICI therapy has revolutionized treatment for advanced melanoma; however, only a subset of patients benefit from this treatment. Despite considerable efforts, the Tumor Mutation Burden (TMB) is the only FDA-approved biomarker in melanoma. However, the mechanisms underlying TMB association with prolonged ICI survival are not entirely understood and may depend on numerous confounding factors,” wrote the investigators.

“To identify more interpretable ICI response biomarkers based on tumor mutations, we train classifiers using mutations within distinct biological processes. We evaluate a variety of feature selection and classification methods and identify key mutated biological processes that provide improved predictive capability compared to the TMB. The top mutated processes we identify are leukocyte and T-cell proliferation regulation, which demonstrate stable predictive performance across different data cohorts of melanoma patients treated with ICI.”

“This work aims to identify better and more biologically interpretable genomic predictors for immunotherapy responses,” noted Noam Auslander, PhD, assistant professor in the molecular cellular oncogenesis program. “We need better biomarkers to help select patients that are more likely to respond to ICI therapy and understand what factors can help to enhance responses and increase those numbers.”

Using machine learning and publicly available de-identified clinical data, the team investigated why some melanoma patients responded to ICI therapy and others did not. Andrew Patterson, a grad student and first author on the paper, details that their research process involved training machine learning models on a dataset to predict whether a patient responds to ICI therapy, and then confirming that the model was able to continually predict response or resistance to this treatment over multiple other datasets.

The team found that leukocyte and T-cell proliferation regulation processes have some mutated genes that contribute to ICI treatment response and resistance. This knowledge could be used to identify targets to enhance responses or mitigate resistance in patients with melanoma.

“We were able to better predict if a patient would respond to ICI therapy than the current clinical standard method as well as extract biological information that could help in further understanding the mechanisms behind ICI therapy response and resistance,” Patterson explained.

The scientists intend to continue this work with the goals of increasing prediction accuracy, further understanding biological mechanisms underpinning patient resistance or responsiveness to ICI therapy, and determining whether the processes distinguished in the paper can also serve as predictors of ICI treatment response for other cancer types.

Also of Interest