From sequence to function through structure: Deep learning for protein design

Ferruz, Noelia; Heinzinger, Michael; Akdel, Mehmet; Goncearenco, Alexander; Naef, Luca; Dallago, Christian

doi:10.1016/j.csbj.2022.11.014

Cited by 55 publications

(49 citation statements)

References 122 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…More specifically, we used a recently published pipeline that combines a variety of protein property predictors that all rely on the pLM ProtT5 to encode protein sequences. This enables efficient processing as the heavy-lifting of generating embeddings needs to be performed only once while the predictors trained add only insignificant computational overhead allowing to predict thousands of proteins within few minutes [69]. Here, we mainly focused on predicting whether a protein is secreted or not using a method called LightAttention and on predicting transmembrane regions using the method TMbed [70].…”

Section: Methodsmentioning

confidence: 99%

Domain loss enabled evolution of novel functions in a gene superfamily

Koludarov

Heinzinger

Jackson

et al. 2022

Preprint

Self Cite

View full text Add to dashboard Cite

3-finger toxins (3FTxs) are a functionally diverse family of toxins, seemingly unique to the venoms of caenophidian snakes. Although the ancestral function of 3FTxs is antagonism of nicotinic acetylcholine receptors (nAChR), redundancy conferred by accumulation of duplicate genes has facilitated extensive neofunctionalization, such that derived members of the family interact with a range of targets. 3FTxs are members of the Ly6/uPAR family, but their precise non-toxin ancestor remains unknown. Using a combination of traditional phylogenetic approaches, manual synteny analysis, and cutting-edge machine learning techniques, we have reconstructed a detailed evolutionary history of 3FTxs. We identify their immediate ancestor as a non-secretory Ly6, unique to squamate reptiles, and propose that changes in molecular ecology resulting from loss of a transmembrane region and changes in gene expression, paved the way for evolution of one of the most important families of snake toxins.

show abstract

Section: Methodsmentioning

confidence: 99%

Domain loss enabled evolution of novel functions in a gene superfamily

Koludarov

Heinzinger

Jackson

et al. 2022

Preprint

Self Cite

View full text Add to dashboard Cite

show abstract

“…This has been followed by a boom in end-to-end learning approaches on proteins sequences for function prediction, as well as on protein structures for generating designed protein sequences. See [223] for a recent review.…”

Section: Challenges and Future Directionsmentioning

confidence: 99%

Beyond sequence: Structure-based machine learning

Durairaj¹,

Ridder²,

Dijk³

2023

Computational and Structural Biotechnology Journal

View full text Add to dashboard Cite

“…They have also been employed to understand the language of nucleic acids (DNA/RNA) and proteins. Protein language models (PLMs), using amino acid alphabets, are the most heavily investigated biological LLMs (Elnaggar et al, 2022, Rives et al, 2021), with demonstrated success in downstream tasks such as protein function prediction (Unsal et al, 2022) and engineering (Ferruz et al, 2022). Nucleotide LLMs, using DNA/RNA alphabets, are still understudied (Avsec et al, 2021).…”

Section: Current State Of the Artmentioning

confidence: 99%

GenSLMs: Genome-scale language models reveal SARS-CoV-2 evolutionary dynamics

Zvyagin

Brace

Hippe

et al. 2022

Preprint

View full text Add to dashboard Cite

Our work seeks to transform how new and emergent variants of pandemic causing viruses, specially SARS-CoV-2, are identified and classified. By adapting large language models (LLMs) for genomic data, we build genome-scale language models (GenSLMs) which can learn the evolutionary landscape of SARS-CoV-2 genomes. By pre-training on over 110 million prokaryotic gene sequences, and then finetuning a SARS-CoV-2 specific model on 1.5 million genomes, we show that GenSLM can accurately and rapidly identify variants of concern. Thus, to our knowledge, GenSLM represents one of the first whole genome scale foundation models which can generalize to other prediction tasks. We demonstrate the scaling of GenSLMs on both GPU-based supercomputers and AI-hardware accelerators, achieving over 1.54 zettaflops in training runs. We present initial scientific insights gleaned from examining GenSLMs in tracking the evolutionary dynamics of SARS-CoV-2, noting that its full potential on large biological data is yet to be realized.

show abstract

From sequence to function through structure: Deep learning for protein design

Cited by 55 publications

References 122 publications

Domain loss enabled evolution of novel functions in a gene superfamily

Domain loss enabled evolution of novel functions in a gene superfamily

Beyond sequence: Structure-based machine learning

GenSLMs: Genome-scale language models reveal SARS-CoV-2 evolutionary dynamics

Contact Info

Product

Resources

About