UniRep50 model from the George Church group (https://www.nature.com/articles/s41592-019-0598-1)
Originally cloned from: https://github.com/churchlab/UniRep The authors of that repository are Ethan Alley, Grigory Khimulya, Surojit Biswas.
Code simplified and modified to carry out batch inference by Martin Engqvist
Assuming that you use Miniconda (https://docs.conda.io/en/latest/miniconda.html) or Anaconda (https://www.anaconda.com/) the required packages can be easily installed from the yml file packages.yml. In a terminal execute:
conda env create -f packages.yml
conda activate tf1.15Download repository and unzip (alternatively fork or clone), cd to the project base folder and execute the command below:
pip3 install -e .If using an anaconda environment you may have to first locate the miniconda pip using whereis.
whereis pipLocate the appropriate file path (the one that has anaconda and the correct environment in the filepath) and run the modified command. For example:
/home/username/miniconda3/envs/env_name/bin/pip install -e .The library should now be available for loading in all your python scripts.
The sequences you wish to convert to UniRep embeddings need to be collected in a single FASTA file. The three embedding types are appended to create a vector of 5700 numbers and are output to a tab-separated file.
>>> from unirep.run_inference import BatchInference
>>> inf_obj = BatchInference(batch_size=256)
>>> df = inf_obj.run_inference(filepath='my_sequences.fasta')
>>> df.to_csv('my_sequences_embeddings.tsv', sep='\t')Copyright 2020-2021 Martin Engqvist
All the model weights are licensed under the terms of Creative Commons Attribution-NonCommercial 4.0 International License. To view a copy of this license, visit http://creativecommons.org/licenses/by-nc/4.0/ or send a letter to Creative Commons, PO Box 1866, Mountain View, CA 94042, USA.
Otherwise the code in this repository is licensed under the terms of GPL v3 as specified by the gpl.txt file.