TX
Free
FEATURED
TxGemma
AI models for accelerating therapeutic development.
A collection of open models designed to predict properties of therapeutic entities such as small molecules, chemicals, and proteins to improve the efficiency of drug discovery and development.
PROS
- + Open-weight and free to use and adapt for research and development. Strong performance across a broad range of therapeutic tasks (66 benchmarks). Includes conversational variants (TxGemma-Chat) to explain reasoning. Versatile capabilities: classification
- + regression
- + and generation tasks. Efficient LLM suitable for data-limited fine-tuning applications.
CONS
- - Requires technical expertise and computational resources for deployment and fine-tuning. Conversational variants have slightly reduced raw prediction performance. Requires accepting the Health AI Developer Foundation's terms of use to access.
-
Predicting a molecule's blood-brain barrier permeability (Classification)
-
Estimating a drug's binding affinity (Regression)
-
Inferring chemical reactants from a reaction product (Generation)
-
Engaging in conversational AI to get mechanistic reasoning for predictions
-
Fine-tuning a foundation model on proprietary therapeutic data
