Early Prediction of Anti-PD-1 Therapy Response in Hepatocellular Carcinoma Using Gut Microbiota Biomarkers

Hepatocellular carcinoma (HCC) remains a leading cause of cancer-related mortality, and response rates to anti-PD-1 therapy are suboptimal. Previous models predicting therapy response often fail to provide reliable early predictions, limiting their clinical utility. To develop a robust predictive model based on gut microbiota features that can accurately discriminate between responders (R) and non-responders (NR) to anti-PD-1 therapy at baseline, enhancing early clinical decision-making. We reanalyzed microbiome abundance data from a cohort of HCC patients receiving anti-PD-1 therapy, originally published by Wu et al. ¹ . By incorporating microbial taxa from phylum to genus levels, we addressed taxonomic assignment uncertainty. Simultaneously, we leveraged low-abundance, informative features via data transformation. Using statistical and machine learning tools, we rectified the informative features, reducing them to 8 key taxa. A neural network model was developed and performance-assessed. Our predictive model, utilizing 8 key microbial taxa, demonstrated an AUROC of 0.784 ± 0.024 on test data. Also, unlike the negative result of the original publication, our model could also successfully discriminate responders from non-responders at baseline, providing a critical advantage at the early intervention stage. This study presents a clinically valuable predictive model that enhances the ability to forecast responses to anti-PD-1 therapy in HCC patients using gut microbiota features. The model’s simplicity and high predictive accuracy offer advantages in personalized treatment planning, contributing to the improvement of patient outcomes in HCC immunotherapy.