DeepGeni: deep generalized interpretable autoencoder elucidates gut microbiota for better cancer immunotherapy

Area of Science:

• Microbiome research
• Cancer immunotherapy
• Computational biology

Background:

• Gut microbiota significantly influences cancer immunotherapy outcomes.
• Fecal microbiota transplantation shows clinical benefits for melanoma patients.
• Identifying specific microbial contributors to treatment response is vital for precision oncology.

Purpose of the Study:

• To develop a robust and interpretable model for predicting cancer immunotherapy response based on microbiome data.
• To enhance the generalizability of machine learning models in microbiome analysis.
• To identify key microbial taxa associated with treatment efficacy.

Main Methods:

• Introduction of DeepGeni, a deep generalized interpretable autoencoder.
• Data augmentation techniques within the autoencoder framework.
• Development of interpretable links to identify informative microbial features.

Main Results:

• DeepGeni-based classifier surpassed state-of-the-art methods in predicting melanoma patient response to immune checkpoint inhibitors.
• The model demonstrated improved generalizability and interpretability of microbiome profiles.
• Key microbial taxa associated with immunotherapy response were elucidated.

Conclusions:

• DeepGeni offers a powerful tool for microbiome-driven prediction of cancer immunotherapy response.
• The findings highlight the potential of targeting specific gut microbes to improve treatment outcomes.
• This study paves the way for microbiome-based therapeutic strategies, including fecal microbiota transplantation and probiotics.