Request for Guidance on Downloading L1000 Signature Data for Multiple FDA-Approved Drugs

[Elmezayen, Ammar - (elmezayen)]

Dear iLINCS Team,

I hope this email finds you well.

My name is Dr. Ammar Elmezayen, and I’m a computational biologist working on a project that involves large-scale analysis of L1000 gene expression signatures for FDA-approved drugs. I am reaching out to seek your guidance on the most efficient way to programmatically access and download these signature files from the iLINCS portal.

🔍 Objective

I aim to download the .xls signature files for a list of FDA-approved drugs using L1000 data from the following URL:

https://www.ilincs.org/ilincs/signaturesL1000/LDG-1188/search/

The selection criteria for each drug are as follows:

One signature per organ

Prefer 10 µM concentration; if unavailable, choose the closest concentration

If multiple cell lines exist per organ, select only one

Prefer 24h time point, unless is_exemplar = 1

Download the signature .xls file.

📄 Current Workflow

Currently, I manually search for each drug under the "Perturbagen" filter at the following URL:

https://www.ilincs.org/ilincs/signaturesL1000/LDG-1188/search/

I then apply the above selection criteria and manually download the .xls. This process is very time-consuming when applied to hundreds of drugs.

🤖 What I’ve Tried

I attempted to automate the download process using:

The iLINCS API (via signature IDs): many returned 404 errors

Selenium browser automation: unable to trigger the Perturbagen filter or reliably select/filter results

Due to the dynamic and complex nature of the UI and JavaScript filtering, these approaches have not been successful.

I would greatly appreciate your advice on any of the following:

1. Is there an API endpoint or backend query method to retrieve signature files using drug names as input?

2. Is there a bulk download option for signatures matching custom criteria (e.g., concentration, cell line, time point)?

3. If automation is the only option, do you recommend a better way to interact with the dynamic filter (e.g., browser APIs, prebuilt query URLs, or parameterized endpoints)?

I’d be happy to provide the full list of drug names I’m working with if that helps.

Thank you very much for your time and for maintaining such a valuable resource. I look forward to your guidance.

Best regards,

Ammar Elmezayen

Computational Biologist