How do i use your code

[RishalAggarwal]

Hi I would like to use ECLAIR to make protein-protein interface residue predictions on a dataset I have, can you guid me on how to use your code for it.

[RishalAggarwal]

guide*

[ShayneWierbowski]

Hi Rishal. I finished my grad work and left the Yu Lab about a year ago, but was the maintainer of ECLAIR related resources prior to that and still get notifications for this repository. I’m not sure who if anyone currently in the lab is receiving these notifications and can comment.

This is a complicated question. The short answer is you don’t. As I recall this repository was only to share code for the model architecture and hyperprameter tuning for the InteractomeINSIDER publication, but did not share pre-trained models, training datasets, or code for feature generation.

Feature generation for this task can be fairly intensive, has many interconnected dependencies, and at the time of publication the code was not sufficiently compartmentalized to extract from the Yu lab’s computer servers to externalization on GitHub.

I would refer you to the lab’s Pioneer Repository instead. This (currently) pre-publication code has been made public and handles an updated model for interface prediction. I believe feature generation steps were incorporated into this repository. I’m not sure if pre-trained model weights are / will be provided (probably not at least until publication if ever?). There’s also a link to a new web portal that may be useful here.

Reach out and open an issue there if the code is not self-explanatory. I believe Dapeng has been managing that repository and should be able to respond to the issue. The Yu lab also has the capacity to generate predictions on new interactions using the old ECLAIR model on a one-off basis, but I don’t know recall who is responsible for that these days.