LINKER-Pred2

A Disordered Flexible Linker (DFL) ppredictor trained on DLD domain Linker dataset and Disprot Database.

📝 Description

LINKER-Pred2 project belongs to a series of LINKER-Pred-Suite projects which focus on the Structure and Function prediction of Disordered Flexible Linkers (DFLs). Currently we have LINKER-Pred2 for fast DFL prediction and LINKER-Pred2 for more accurate DFL prediction. LINKER-Pred2 is trained on more than 2000 DFL linker dataset from our DLD dataset and Disprot, its performance is better than TOP predictors on the CAID2 Linker dataset.

🐣 Getting Started

Currently, we provide two ways to use this predictor: Docker or download the source code from GitHub.

Pre-requirements

This predictor requires sequences embedded with ProtTrans and MSA Transformer. Note,

• File format should be [SEQUENCE_NAME/ID].npy, replace SEQUENCE_NAME/ID with the actual sequence ID, it should be the same as the name from the .fasta file.
• Matrix shape:
  Onehot: (1, SEQUENCE_LENGTH, 21)
  ProtTrans: (1, SEQUENCE_LENGTH, 1024)
  MSA Transformer: (1, SEQUENCE_LENGTH, 768)

📣‼️If you don't have them available, please visit the embedding section of our project first to embed the sequences.‼️

(We maintain this separation due to the requirements from CAID3, but we may edit or merge them in the future.)

Docker (Recommended)

Dependencies

• Go to embedding if you don't have ProtTrans and MSA Transformer embedded sequences;
• Docker Desktop 4.27.2 or higher;

Installing

• Pull the Docker image from DockerHub

  docker pull dimeng851/punch_linker2:v2

Executing program

• RUN the following command:

  Replace
  CONTAINER_NAME - any name you like;
  PATH_TO_INPUT_FASTA - path to input file, which is ONE FASTA file including all query sequences;
  PATH_TO_MSATRANS - a folder which includes all MSA Transformer embedded sequences;
  PATH_TO_PROTTRANS - a folder which includes all protTrans embedded sequences;
  PATH_OUTPUT - a folder which will be used to save all outputs, including: a. timings.csv; b. disorder folder, where will keep all the prediction resulds.

  docker run -d \
  -it \
  --name [CONTAINER_NAME] \
  --mount type=bind,source=[PATH_TO_INPUT_FASTA],target=/punch_linker2/data/input.fasta \
  --mount type=bind,source=[PATH_TO_MSATRANS],target=/punch_linker2/data/msaTrans \
  --mount type=bind,source=[PATH_TO_PROTTRANS],target=/punch_linker2/data/protTrans \
  --mount type=bind,source=[PATH_OUTPUT],target=/punch_linker2/output \
  dimeng851/punch_linker2:v2

  An example:

  docker run -d \
  -it \
  --name punch_linker2_con \
  --mount type=bind,source=data/input.fasta,target=/punch_linker2/data/input.fasta \
  --mount type=bind,source=data/msaTrans,target=/punch_linker2/data/msaTrans \
  --mount type=bind,source=data/protTrans,target=/punch_linker2/data/protTrans \
  --mount type=bind,source=data/results,target=/punch_linker2/output \
  dimeng851/punch_linker2:v2

• Find the results in OUTPUT folder.

Contact & Help 📩

Email Di.

di.meng@ucdconnect.ie

Authors

📬 Di Meng - di.meng@ucdconnect.ie
📬 Juliana Glavina - jglavina@iib.unsam.edu.ar
📬 Gianluca Pollastri - gianluca.pollastri@ucd.ie
📬 Lucía Beatriz Chemes - lchemes@iib.unsam.edu.ar

Project

https://github.com/deemeng/punch_linker2