UniRep

UniRep is a deep learning-based software tool that generates a unified statistical representation of proteins from their amino acid sequences. This representation captures fundamental protein features grounded in structure, evolution, and biophysical properties. The key aspects of UniRep are:

It learns from unlabeled amino acid sequences, allowing it to distill the essential features of a protein without requiring additional information.

The generated representation is semantically rich and encodes meaningful information about the protein's properties and function.

Models built on top of UniRep are widely applicable and can generalize to previously unseen protein sequences.

UniRep can predict the stability of natural and de novo designed proteins and the quantitative function of diverse protein mutants, with performance comparable to state-of-the-art methods.

Topic

Protein folding, stability and design;Informatics;Machine learning;Synthetic biology;Biophysics

Detail

  • Operation: Ab initio structure prediction;Protein quantification;Protein function prediction

  • Software interface: Command-line user interface

  • Language: Python

  • License: Not stated

  • Cost: Free of charge

  • Version name: -

  • Credit: NIH, NSF GRFP Fellowship, NIGMS, the Center for Effective Altruism, the Wyss Institute for Biologically Inspired Engineering.=

  • Input: -

  • Output: -

  • Contact: George M. Church gchurch@genetics.med.harvard.edu

  • Collection: -

  • Maturity: -

Publications

  • Unified rational protein engineering with sequence-based deep representation learning.
  • Alley EC, et al. Unified rational protein engineering with sequence-based deep representation learning. Unified rational protein engineering with sequence-based deep representation learning. 2019; 16:1315-1322. doi: 10.1038/s41592-019-0598-1
  • https://doi.org/10.1038/S41592-019-0598-1
  • PMID: 31636460
  • PMC: PMC7067682

Download and documentation

< Back to DB search