Evaluation of a novel signature for PARP inhibitor sensitivity prediction using real-world data.
Evaluation of a novel PARP inhibitor sensitivity biomarker using real-world data.
First Author: Emily Vucic, Zephyr AI
Background
Tumors with homologous recombination repair deficiencies (HRD+) are susceptible to PARP inhibitors (PARPi) which exploit a synthetic lethality by targeting an essential base excision repair (BER) pathway. Current biomarkers designed to predict efficacy of PARPi (BRCA1/2 mut, HRD+) are ineffective, particularly in advanced stages of therapy. There is an urgent unmet need for improved PARPi biomarkers that is particularly acute for patients who fail multiple lines of therapy. We sought to develop and validate a predictive signature for identifying olaparib-sensitive patients that uses real-world data (RWD) as input.
Research Background
Tumors with homologous recombination repair deficiencies (HRD+) are susceptible to PARP inhibitors (PARPi) which exploit a synthetic lethality by targeting an essential base excision repair (BER) pathway. Current biomarkers designed to predict efficacy of PARPi (e.g., BRCA1/2 mutations, HRD+) are ineffective, particularly in advanced stages of therapy. There is an urgent unmet need for improved PARPi biomarkers that is particularly acute for patients who fail multiple lines of therapy. This study aimed to develop and validate a predictive signature for identifying patients sensitive to olaparib, using real-world data (RWD) as input.
Methods
We developed a machine learning (ML) model, called Drug-BERT, that learns a relationship between 1) drug structure, 2) DNA alterations, 3) drug response and 4) binding affinity between a drug and its purported target. Drug BERT utilizes genetic information from clinical commercial NGS panels and a subset of clinical features, and outputs a predictive signature for a drug of interest, without need for additional training. Each prediction is accompanied by a Vulnerability Network (VN) which represents a predicted latent genetic sensitivity in tumors and provides biological interpretability to model outputs. Because our model uses RWD as input, we were able to retrospectively evaluate our signature in a RWD ovarian cancer cohort treated with the PARPi, olaparib. Samples were filtered to biopsies taken 2 years prior to olaparib treatment (N= 48 samples). Drug combinations were analyzed by processing each drug separately and stratifying patients based on sensitivity predictions to all drugs in the regimen.
Research Methods
This study developed a machine learning model called Drug-BERT, which learns the relationships among four aspects: 1) drug structure; 2) DNA alterations; 3) drug response; and 4) binding affinity between a drug and its target. Drug-BERT utilizes genetic information from clinical NGS panels and a subset of clinical features, outputting a predictive signature for the target drug without additional training. Each prediction is accompanied by a Vulnerability Network (VN), representing the predicted latent genetic sensitivity in tumors and providing biological interpretability to the model outputs. Since the model uses RWD as input, we could retrospectively evaluate the predictive signature in a RWD ovarian cancer cohort treated with the PARPi olaparib. Samples were filtered from biopsies taken within two years prior to the start of olaparib treatment (N=48). The analysis of drug combinations was performed by processing each drug separately and stratifying patients based on sensitivity predictions for all drugs in the regimen.
Results
Zephyr’s signature outperformed existing PARPi biomarkers in a complex real-world olaparib-treated ovarian cancer cohort. OS and PFS were not significantly different when stratifying patients by BRCA1/2 mutations or HRD+. For Zephyr’s signature, both OS (p-value<5×10-3; HR = 3.37, 95% CI: 1.27-5.46) and PFS (p-value<10-3; HR = 2.04, 95% CI: 1.35-3.10), were statistically and clinically significantly prolonged. Characterization of VNs enriched in olaparib-predicted sensitive patient tumors indicated enhanced sensitivity to perturbing HR and cell cycle pathways. Conversely, olaparib predicted non-sensitive ovarian tumors were characterized by VNs indicating perturbation sensitivity to cell motility and stress response pathways and displayed higher expression of DNA repair, drug resistance, and cancer stem cell pathways.
Research Results
In a complex real-world cohort of ovarian cancer patients treated with olaparib, the predictive biomarker generated by Zephyr AI outperformed existing PARPi biomarkers. There were no significant differences in overall survival (OS) and progression-free survival (PFS) when stratifying patients by BRCA1/2 mutations or HRD+. However, for the predictive biomarker from Zephyr AI, both OS (p<0.005; HR=3.37, 95% CI 1.27-5.46) and PFS (p<0.001; HR=2.04, 95% CI 1.35-3.10) were statistically and clinically significantly prolonged. The characterization of VNs enriched in tumors predicted to be sensitive to olaparib indicated enhanced sensitivity to perturbations in homologous recombination repair and cell cycle pathways. Conversely, tumors predicted to be non-sensitive to olaparib exhibited VNs indicating sensitivity to perturbations in cell motility and stress response pathways and showed higher expression of DNA repair, drug resistance, and cancer stem cell pathways.
Conclusions
We present a novel and enhanced approach for stratifying patients for PARPi therapy, retrospectively validated in a complex real-world ovarian cancer cohort treated with olaparib. Our interpretable model not only offers a biological hypothesis for predicted responses but is also readily applicable in clinical settings and compatible with standard multi-gene commercial NGS panels for immediate use.
Research Conclusions
This study developed a novel and optimized method for stratifying patients receiving PARPi therapy, which was retrospectively validated in a complex real-world ovarian cancer cohort treated with olaparib. Our interpretable model not only provides a biological hypothesis for efficacy prediction but is also applicable in real clinical settings and compatible with existing commercial standard NGS multi-gene panels.
Professor Lin Zhongqiu’s Commentary:
BRCA1/2 and HRD testing remain essential indicators for assessing PARPi efficacy and guiding treatment. Other predictive indicators only become meaningful when PARPi resistance occurs. The AI model mentioned in this article is similar.
Editor: Li Ruixin
Reviewer: Lin Zhongqiu