Spaces:
Paused
Apply for a GPU community grant: Academic project
Hello Hugging Face team,
I am Md Delower Hossain, a doctoral researcher and PhD candidate in AI at the University of Alabama at Birmingham (UAB), working with the SPARC team on computational drug discovery. I am applying for a ZeroGPU community grant for this Space, Delower/dpp4-kd-webapp.
Description of the app. This Space is an interactive binding-affinity (Kd) predictor for DPP-4 (dipeptidyl peptidase-4), the type-2-diabetes drug target. It serves our DPP-4 Kd LTN model โ the best model of an open transfer-learning suite we built for this target. The model is a Logic Tensor Network (MLP trunk 768-384-192) over Morgan fingerprints plus the ESM-2 protein embedding of human DPP-4 (UniProt P27487). It was pre-trained on an all-target Kd corpus (86,569 compound-target rows) and then fine-tuned on DPP-4 Ki data (555 compounds); on the 15 DPP-4 Kd compounds that were held out of every training stage it scores RMSE 1.604 / MAE 1.276 (pKd) with 0.867 potency-class accuracy. Visitors enter a SMILES, singly or as a batch CSV, and receive a predicted pKd, Kd in nM, ฮG and potency class. The app is intended for researchers and students exploring DPP-4 inhibitor candidates, as a companion to the Kd values we compute for the candidate molecules from our generative DPP-4 pipeline. The standalone model, with its config and evaluation, is published at Delower/dpp4-kd-ltn, and the Space's complete app code is public in this repo.
Justification. The project is fully open and non-commercial. My account's two free ZeroGPU Spaces are already used by the earlier apps in this series โ Delower/dpp4-pic50-webapp (DPP-4 pIC50 predictor) and Delower/egfr-pic50-webapp (EGFR pIC50 predictor), both live on ZeroGPU โ and creating a third Gradio Space now requires a paid plan, which is why I am applying through the community grant programme. One point of transparency: unlike a large generative model, this predictor is a small MLP (~3M parameters) and answers in milliseconds, so it would also run acceptably on CPU. I am requesting ZeroGPU because it keeps this app in the same free, open research-app family as its two companions, because each visitor's usage then counts against their own ZeroGPU quota, and because the grant is the mechanism HF provides for a third Gradio Space on a free account. If the reviewers would prefer to attach a different hardware tier, I am glad to accept whatever they judge appropriate.
The Space currently exists as a static placeholder with the complete, locally verified Gradio app already uploaded; I will flip it fully live as soon as the grant is attached. I am happy to provide any further context, and I will credit Hugging Face ZeroGPU in the Space and in the project documentation. Thank you very much for supporting open research demos.
โ Md Delower Hossain (Delower), UAB SPARC