Background A high-quality sequence alignment (SA) is the most important input feature for accurate protein structure prediction. For a protein sequence, there are many methods to generate a SA. However, when given a choice of more than one SA for a protein sequence, there are no methods to predict which SA may lead to more accurate models without actually building the models. In this work, we describe a method to predict the quality of a protein's SA.
Methods We created our own dataset by generating a variety of SAs for a set of 1,351 representative proteins and investigated various deep learning architectures to predict the local distance difference test (lDDT) scores of distance maps predicted with SAs as the input. These lDDT scores serve as indicators of the quality of the SAs.
Results Using two independent test datasets consisting of CASP13 and CASP14 targets, we show that our method is effective for scoring and ranking SAs when a pool of SAs is available for a protein sequence. With an example, we further discuss that SA selection using our method can lead to improved structure prediction.
http://deep.cs.umsl.edu/sa-scoring/
http://deep.cs.umsl.edu/sascoring/download
- Download the repository
git clone https://github.com/ba-lab/Alignment-Score.git
OR
Download the zip file: https://github.com/ba-lab/Alignment-Score/archive/refs/heads/main.zip
cd Alignment-Score
- Download additional model file
wget http://deep.cs.umsl.edu/sascoring/download/deep-msa-score/trRosetta_model2019_07.tar.xz
tar -xf trRosetta_model2019_07.tar.xz
- Install virtual environments using conda
conda env create -f trRos-env.yml
conda env create -f msascore-env.yml
Instructions to install conda can be found here: https://varhowto.com/install-miniconda-ubuntu-20-04/ OR, Official site: https://docs.conda.io/en/latest/miniconda.html
chmod +x automating_script.sh
./automating_script.sh <path_to_msa> <target_name>
Eg command:
./automating_script.sh test_msa/T1024_DeepMSA2.a3m T1024
Bioinformatics, 2022, Scoring protein sequence alignments using deep learning