aerijman / adpred-new-website Goto Github PK

IN the website add the residue type on the mouse hover

Tasmanian should throw a comment when there is not enough reads to compute the mode of the read_length

1. when you get the email with the results, it will be super helpful to know what protein was searched. Is there some way to put the protein name (uniprot ID) and/or the fasta name in results files as we discussed yesterday?

2. We should make it clear that this predictor seems specific for “acidic” activation domains. I suggest the following changes to the homepage. Hopefully, you have permission to do text changes to the homepage yourself without Dan’s help.

Title bar and text recommended changes (altered text in magenta):
ADpred

Prediction of acidic tanscription activation domains (ADs) from protein sequences

What is ADpred?

ADpred is a deep learning model to predict acidic transcription activation domains (ADs) within protein sequences. It has been described in Erijman et al. from the Hahn lab and the Söding lab

How does the ADpred webserver work?

A protein sequence OR protein ID can be provided. The protein ID is used to retrieve the sequence from uniprot, hence, make sure that your ID is unique and does not expand to different organisms. The secondary structure is retrieved from the psipred webserver and both sequence and secondary structure are the input of ADpred. ADpred uses a convolutional deep neural network to predict the probability of peptides having potential AD function. The practical length limits for ADs identified by ADpred is between ≥9 to ≤30 residues. We typically look for ADpred scores of ≥0.8 over ≥10-15 continuous residues as indicating a high propensity for AD function. However, we have found that not all peptides with potential AD function are in the proper protein context to work as transcription activators. See Erijman et al for details.

Once the results are ready, you should receive an email with a link to a file with the results.

Questions or comments to: aerijman@xxx