code and data for S2asolP

• Install Anaconda (https://www.anaconda.com/download) create S2asolP environment. (conda env create -f s2asolp.yaml)

• Install SCRATCH-1D^[1] release 1.2 (https://download.igb.uci.edu/SCRATCH-1D_1.2.tar.gz)

• R requirements (https://www.r-project.org)

  • R libraries
    • bio3d
    • stringr
    • Interpol
    • zoo

• Download the sa-prot^[2] model as encoder from https://huggingface.co/westlake-repl/SaProt_650M_PDB and place it into model folder. You can create bio environment by conda. (conda env create -f bio.yaml) Use conda activate s2asolp or conda activate R to activate the environment.

We have placed the computed results in the infer_res folder.

1. Download S2asolP data and model in https://drive.google.com/drive/folders/1SqC5NWzTx_McoL9l6KlUwWYmon8E-4mF?usp=sharing
2. Then unzip the downloaded data and place it into the data folder, and move s2asolp_checkpoint.pt into the checkpoints folder.
3. Activate conda env (source activate s2asolp).
4. Run the bash_infer.sh (Bash bash_infer.sh).

You need to perform the following steps to predict new test file (e.g. test_seq.fasta).

• Run SCRATCH with the new test file.
  • Execute in the command line: Run your_SCRATCH_installation_path/bin/run_SCRATCH-1D_predictors.sh test_seq.fasta test_seq 8 8 is the number of processors, test_seq is the output files' prefix.
  • It will return four files in current folder:
    • test_seq.ss
    • test_seq.ss8
    • test_seq.acc
    • test_seq.acc20
• Calculate features for test sequences.
  • Execute in the command line: Run R --vanilla < PaRSnIP.R test_seq.fasta test_seq.ss test_seq.ss8 test_seq.acc20 test_seq
  • After this step, one file will be created:
    • test_seq_src_bio: contains biological features corresponding to the raw protein sequences
• Use AlphaFold^[3] or ColabFold^[4] to get test sequences' pdb file
• Use Foldseek^[5] to get the test sequences' 3di file
• Run get_3di.py to get input sequence
• Replace the parameters in bash_infer.shand run the script Bash bash_infer.sh to infer the test sequences result, or replace the parameters in bash_s2asolp.sh and run the script Bash bash_s2asolp.sh to retrain the model.

1. Magnan C N, Baldi P. SSpro/ACCpro 5: almost perfect prediction of protein secondary structure and relative solvent accessibility using profiles, machine learning and structural similarity[J]. Bioinformatics, 2014, 30(18): 2592-2597.

2. Su J, Han C, Zhou Y, et al. SaProt: protein language modeling with structure-aware vocabulary[J]. bioRxiv, 2023: 2023.10. 01.560349.

3. Jumper J, Evans R, Pritzel A, et al. Highly accurate protein structure prediction with AlphaFold[J]. Nature, 2021, 596(7873): 583-589.

4. Mirdita M, Schütze K, Moriwaki Y, et al. ColabFold: making protein folding accessible to all[J]. Nature methods, 2022, 19(6): 679-682.

5. Van Kempen M, Kim S S, Tumescheit C, et al. Fast and accurate protein structure search with Foldseek[J]. Nature Biotechnology, 2023: 1-4.