IDPFunNet(A novel deep learning framework for systematic prediction of intrinsically disordered protein functions using protein language model) is a software designed for the accurate prediction of various disordered binding functions and disordered flexible linker functions of proteins. Due to the different characteristics of disordered binding functions and flexible linker functions, these two categories are predicted separately. Using sequence semantic vectors obtained from the protein language model ProtT5, a multi-task learning framework and multi-scale information fusion network are employed to predict disordered functional annotations for protein binding, nucleic acid binding, lipid binding, ion binding, and small molecule binding. A global information network is used to predict the disordered functional annotations of flexible linker.
For more details on IDPFunNet, please refer to the manuscript.
Step 1: Install Anaconda
The anaconda can be downloaded from https://www.anaconda.com/
Step 2: Clone this repository
Now, you need to clone this Github repository with
git clone https://github.com/IDRIDP/IDPFunNet.git
Step 3: Download the ProtT5 network weights
Download the ProtT5 model weights from the following link and copy the file to the "/IDPFunNet/utils/prot_t5_xl_uniref50/" directory.
https://huggingface.co/Rostlab/prot_t5_xl_uniref50/resolve/main/pytorch_model.bin
Step 4: Navigate to the installation path
cd IDPFunNet
Step 5: Create a new environment
conda env create -f IDPFunNet.yml
conda activate IDPFunNet
1 Predict disordered multiple binding functions
python predict.py -t binding -i example.fasta -d gpu0
The files binding_scores.txt and binding_binary.txt are saved in the directory "/IDPFunNet/data_save/example/result/". binding_scores.txt contains the predicted scores, and binding_binary.txt contains the binary classification results for disordered binding functions of each sequence.
File Contents:
Line 1: >Sequence ID
Line 2: Protein sequence (1-letter amino acid encoding)
Line 3: Predicted results for disordered protein-binding functions
Line 4: Predicted results for disordered nucleic acid-binding functions
Line 5: Predicted results for disordered lipid-binding functions
Line 6: Predicted results for disordered ion-binding functions
Line 7: Predicted results for disordered small molecule-binding functions
2 Predict disordered flexible linker functions
python predict.py -t linker -i example.fasta -d gpu0
The files linker_scores.txt and linker_binary.txt are saved in the directory "/IDPFunNet/data_save/example/result/". linker_scores.txt contains the predicted scores, and linker_binary.txt contains the binary classification results for disordered flexible linker functions of each sequence.
File Contents:
Line 1: >Sequence ID
Line 2: Protein sequence (1-letter amino acid encoding)
Line 3: Predicted results for disordered flexible linker functions
3 Predict disordered multiple binding and flexible linker functions
python predict.py -t all -i example.fasta -d gpu0
The files all_scores.txt and all_binary.txt are saved in the directory "/IDPFunNet/data_save/example/result/". all_scores.txt contains the predicted scores, and all_binary.txt contains the binary classification results for disordered functions of each sequence.
File Contents:
Line 1: >Sequence ID
Line 2: Protein sequence (1-letter amino acid encoding)
Line 3: Predicted results for disordered protein-binding functions
Line 4: Predicted results for disordered nucleic acid-binding functions
Line 5: Predicted results for disordered lipid-binding functions
Line 6: Predicted results for disordered ion-binding functions
Line 7: Predicted results for disordered small molecule-binding functions
Line 8: Predicted results for disordered flexible linker functions
4 Explanation of some parameters
(1) -t specifies the prediction type, with three options: binding, linker, and all, representing disordered multiple binding functions, disordered flexible linker functions, or both, respectively.
(2) -i specifies the input FASTA file.
(3) -d specifies the processor to use, allowing you to choose between gpu or cpu. If you select gpu, you need to append a number to indicate the GPU card index(e.g., gpu0).
Liang et al, Hybrid deep learning with protein language models and dual-path architecture for predicting IDP functions, Briefings in Bioinformatics, 27: bbag126 (2026).