dimeng851/punch_linker2 repository overview

⁠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:v4
  

  Copy

⁠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:v4
  

  Copy

  An example:

  docker run -d \
  -it \
  --name punch_linker2_con \
  --mount type=bind,source=/Users/deemeng/Downloads/data/linker/linker.fasta,target=/punch_linker2/data/input.fasta \
  --mount type=bind,source=/Users/deemeng/Downloads/data/linker/msaTrans,target=/punch_linker2/data/msaTrans \
  --mount type=bind,source=/Users/deemeng/Downloads/data/linker/protTrans,target=/punch_linker2/data/protTrans \
  --mount type=bind,source=/Users/deemeng/Downloads/data/linker/output,target=/punch_linker2/output \
  dimeng851/punch_linker2:v4
  

  Copy
• Find the results in OUTPUT folder.

⁠Contact & Help 📩

Email Di.

[email protected]

⁠Authors

📬 Di Meng - [email protected]⁠
📬 Juliana Glavina - [email protected]⁠
📬 Gianluca Pollastri - [email protected]⁠
📬 Lucía Beatriz Chemes - [email protected]⁠

⁠Project