RT Journal Article
SR Electronic
T1 ProtTrans: Towards Cracking the Language of Life’s Code Through Self-Supervised Learning
JF bioRxiv
FD Cold Spring Harbor Laboratory
SP 2020.07.12.199554
DO 10.1101/2020.07.12.199554
A1 Ahmed Elnaggar
A1 Michael Heinzinger
A1 Christian Dallago
A1 Ghalia Rehawi
A1 Yu Wang
A1 Llion Jones
A1 Tom Gibbs
A1 Tamas Feher
A1 Christoph Angerer
A1 Martin Steinegger
A1 Debsindhu Bhowmik
A1 Burkhard Rost
YR 2021
UL http://biorxiv.org/content/early/2021/05/04/2020.07.12.199554.abstract
AB Computational biology and bioinformatics provide vast data gold-mines from protein sequences, ideal for Language Models taken from NLP. These LMs reach for new prediction frontiers at low inference costs. Here, we trained two auto-regressive models (Transformer-XL, XLNet) and four auto-encoder models (BERT, Albert, Electra, T5) on data from UniRef and BFD containing up to 393 billion amino acids. The LMs were trained on the Summit supercomputer using 5616 GPUs and TPU Pod up-to 1024 cores.Dimensionality reduction revealed that the raw protein LM-embeddings from unlabeled data captured some biophysical features of protein sequences. We validated the advantage of using the embeddings as exclusive input for several subsequent tasks. The first was a per-residue prediction of protein secondary structure (3-state accuracy Q3=81%-87%); the second were per-protein predictions of protein sub-cellular localization (ten-state accuracy: Q10=81%) and membrane vs. water-soluble (2-state accuracy Q2=91%). For the per-residue predictions the transfer of the most informative embeddings (ProtT5) for the first time outperformed the state-of-the-art without using evolutionary information thereby bypassing expensive database searches. Taken together, the results implied that protein LMs learned some of the grammar of the language of life. To facilitate future work, we released our models at https://github.com/agemagician/ProtTrans.Competing Interest StatementThe authors have declared no competing interest.