SPOT-1D-LM: Reaching Alignment-profile-based Accuracy in Predicting Protein Secondary and Tertiary Structural Properties without Alignment.
Hardware Requirements: SPOT-1D-LM predictor has been tested on standard ubuntu 18 computer with approximately 32 GB RAM to support the in-memory operations.
- Python3.7
- Anaconda
- CUDA 10.1 (Optional if using GPU)
- cuDNN (>= 7.4.1) (Optional if using GPU)
To install SPOT-1D-LM and it's dependencies following commands can be used in terminal:
git clone https://github.com/jas-preet/SPOT-1D-LM.gitcd SPOT-1D-LM
To download the model check points from the dropbox use the following commands in the terminal:
wget https://apisz.sparks-lab.org:8443/downloads/Resource/Protein/2_Protein_local_structure_prediction/SPOT-LM-checkpoints.xztar -xvf SPOT-LM-checkpoints.xz
To install the dependencies and create a conda environment use the following commands
conda create -n spot_1d_lm python=3.7conda activate spot_1d_lm
if GPU computer:
7. conda install pytorch==1.7.0 torchvision==0.7.0 cudatoolkit=10.1 -c pytorch
for CPU only
7. conda install pytorch==1.6.0 torchvision==0.7.0 cpuonly -c pytorch
-
pip install fair-esm -
conda install pandas=1.1.1 -
conda install tqdm -
pip install bio-embeddings[all]
To run SPOT-1D-LM use the following command
bash run_SPOT-1D-LM.sh file_lists/test_file_list.txt cpu cpu cpu to run model, ESM-1b and ProtTrans on cpu
or
bash run_SPOT-1D-LM.sh file_lists/test_file_list.txt cpu cpu cuda:0 to run model on gpu and, ESM-1b and ProtTrans on cpu
or
bash run_SPOT-1D-LM.sh file_lists/test_file_list.txt cuda:0 cuda:1 cuda:2 to run model, ESM-1b and ProtTrans on gpu
for more details on this work refer the manuscript
Please also cite and refer to ESM-1b and ProtTrans as the input used in this work is from these works.