Move fast and learn things
Latest publications
Do protein language models understand evolution? Mixed evidence from ancestral sequences and ESM2
Protein language models are trained on evolutionarily related sequences, yet the extent to which they capture the underlying evolutionary relationships remains unclear. We explore this question using reconstructed ancestral protein sequences and the ESM2 protein language model.
Read now
A framework for modeling human monogenic diseases by deploying organism selection
We designed a decision-making framework to find tractable genes from our organism selection dataset for pilot experiments. We focused on genes in two potential models of human monogenic disease, the choanoflagellate Salpingoeca rosetta and the tunicate Ciona intestinalis.
Read now
Modeling human monogenic diseases using the choanoflagellate _Salpingoeca rosetta_
We applied a decision-making framework for identifying tractable genes from our organism selection dataset for pilot experiments in the choanoflagellate Salpingoeca rosetta.
Read now
Modeling human monogenic diseases using the tunicate _Ciona intestinalis_
We applied a decision-making framework for identifying tractable genes from our organism selection dataset for pilot experiments in the tunicate Ciona intestinalis.
Read now
Graph neural networks: A unifying predictive model architecture for evolutionary applications
The transition from explanatory to predictive models in evolutionary biology is a significant and challenging task. We propose that graph representations and graph neural networks may play a crucial role in this transition.
Read now
Phylogenies and biological foundation models
Biological foundation models are, at their core, evolutionary comparisons on massive scales. As with all comparative studies, evolutionary nonindependence determines their power. We chart how this affects biological AI and propose practical routes to set the field on firmer ground.
Read now