TxGemma

About TxGemma

TxGemma is a collection of machine learning models for therapeutic research. Researchers can supply molecular representations, protein sequences, or related task information to obtain classifications, numerical predictions, and generated responses. Supported examples include estimating molecular properties, toxicity, target binding, and relationships between biological entities. The models are trained on therapeutic task data and can be adapted through instruction-based fine-tuning.

Predictive variants focus on task outputs, while conversational variants let users ask follow-up questions about those predictions. Example notebooks show how to run models, fine-tune them, and combine them with other tools in research agents. TxGemma provides reusable components for computational discovery workflows and requires evaluation for each intended research application.