The episode discusses AlphaGenome 2, an AI model from Google DeepMind that reads the human genome to predict the effects of mutations at the cellular level. The model helps prioritize variants in rare diseases and complex conditions, bridging the gap between genotype and phenotype for precision medicine.
Summarized by Podsumo
AlphaGenome 2 predicts the effects of single-letter mutations across 9 billion possible changes in the human genome, down to individual cell types.
The model successfully identified a causal variant in a rare pediatric epilepsy case that had baffled clinical geneticists for years.
Unlike earlier models focused on bulk tissues, AlphaGenome 2 can predict variant effects in specific cell types, such as macrophages in the intestine linked to inflammatory bowel disease.
The AlphaGenome Atlas provides a free, no-code website for researchers to look up variant impact scores and molecular consequences.
The long-term vision is a 'virtual cell' that simulates biological responses to genetic changes or drugs, though this remains a distant goal.
"The fact is, we now have the kind of level of data where deep learning can really take hold. And we've been able to build together the kind of deep learning models that are able to read a DNA sequence, a bit like you might read a piece of text."
"I think our models are high precision, low recall tools—when they predict a strong effect, they're usually right, but there's a lot they miss."
"Biology is very humbling. I'm quietly confident we can get to this vision of a virtual cell in the future, but we need to make decisions now that increase our probability of success."