Identification of predictive subphenotypes for clinical outcomes using real world data and machine learning by Weishen Pan & Deep Hathi & Zhenxing Xu & Qiannan Zhang & Ying Li & Fei Wang instant download

Identification of predictive subphenotypes for clinical outcomes using real world data and machine learning by Weishen Pan & Deep Hathi & Zhenxing Xu & Qiannan Zhang & Ying Li & Fei Wang instant download

Predicting treatment response is an important problem in real-world applications, where the heterogeneity of the treatment response remains a significant challenge in practice. Unsupervised machine learning methods havebeen proposed to address this challenge by clustering patients with similarelectronic health record (EHR) data. However, they cannot guarantee coherentoutcomes within the groups. Here, we propose Graph-Encoded Mixture Survival (GEMS) as a general machine learning framework to identify distinctpredictive subphenotypes that guarantee coherent survival and baselinecharacteristics within each subphenotype. We apply our method to a realworld dataset of advanced non-small cell lung cancer (aNSCLC) patientsreceiving first-line immune checkpoint inhibitor (ICI) therapy to predict overallsurvival (OS). Our method outperforms baseline methods for predicting OSand identifies three reproducible subphenotypes associated with distinctbaseline clinical characteristics and OS. Our results demonstrate that ourmethod can provide insights in the heterogeneity of treatment response andpotentially influence treatment selection.