Hi AlphaGenome team/community,
I am conducting non-commercial academic research on genetic variant analysis and would like to clarify whether the following use of AlphaGenome is permitted under the current Terms of Service.
I am considering extracting internal representations (embeddings) from AlphaGenome and using them as input features for an independently trained downstream machine-learning classifier:
DNA variant/sequence → AlphaGenome embedding → downstream classifier
The downstream model would be trained using independently collected and annotated data for a biological classification task. It would not be trained to reproduce, approximate, or distill AlphaGenome’s genomic track predictions or other predictive outputs.
Could you please clarify:
- Are internal representations/embeddings extracted from AlphaGenome considered AlphaGenome “Output,” “Derivatives,” or other “AlphaGenome Assets” under the current Terms?
- Would it be permitted to use these embeddings as input features for training an independently developed downstream classifier for non-commercial academic research?
The research is intended for scientific publication.
Thank you very much for your clarification.