Reproduce “ProtTrans: Towards Cracking the Language of Life's Code Through Self-Supervised Deep Learning and High Performance Computing” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Ahmed Elnaggar, Michael Heinzinger, Christian Dallago, Ghalia Rihawi, Yu Wang, Llion Jones, Tom Gibbs, Tamas Feher and 4 more
ProtTrans trains language models on unlabeled protein sequences, then uses frozen embeddings from a single sequence with a small CNN to predict secondary structure without an MSA input. On NEW364, ProtT5-XL-U50's Q3 of 84.8% was near NetSurfP-2.0's 84.3%, with a standard error of plus or minus 0.5 percentage points. Whether embeddings implicitly encode evolutionary information remains unresolved.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.