Discrete flow matching enables protein sequence-structure co-generation. arxiv.org/abs/2402.04997

Finetune a protein language model to correctly predict which sequence in a pair has higher fitness. www.biorxiv.org/content/10.1...

Who would've thought the most useful github repo in this subfield of ours would have zero code?

GitHub - yangkky/Machine-learning-for-proteins: Listing of papers about machine learning for proteins.

Listing of papers about machine learning for proteins. - GitHub - yangkky/Machine-learning-for-proteins: Listing of papers about machine learning for proteins.

Nice review of machine learning for enzyme engineering pubs.acs.org/doi/full/10....

Use optimal transport to learn how to convert variable-length protein language model outputs into fixed-length protein-level representations. www.biorxiv.org/content/10.1...

A completely unsupervised way of extracting contact maps from protein language models helps demonstrate that pLMs learn structure by storing motifs of pairwise contacts. @sokrypton.orgwww.biorxiv.org/content/10.1...

Lots of discussion on plagiarism. But to me, it seems that people we should be really trying to punish for this are also the people that you can't detect via simple text matching. I'm talking about "plagiarism of ideas" either from pre-prints or conferences or discussion etc. 🤔

Self-supervised training on RNA sequences enables zero-shot secondary structure prediction including pseudoknots www.biorxiv.org/content/10.1...

Use protein language models to annotate viral proteins with previously unknown functions. @microbegrrl.bsky.socialwww.nature.com/articles/s41...

Use a VAE to generate adenovirus hexon proteins that evade human antibodies after training on only 711 examples. www.nature.com/articles/s42...