Work

DrugForm: Drug Design Platform — Model Weights and Training Data

Year: 2026

Type: dataset

Abstract: This dataset contains supplementary data for the DrugForm drug design platform: pre-trained UniQSAR model weights, encoder checkpoints (ESM2, ChemFormer), Transforcer models, and associated training datasets (BindingDB, KIBA, MoleculeNet PCBA, etc.). The archive is designed to be extracted from the root of a cloned DrugForm repository, placing all required model and data files into their expected locations. DrugForm is a comprehensive software platform with a web interface built on a flexible microservice architecture, integrating physicochemical property calculation tools, automated predictive model training (UniQSAR framework), and de novo molecular structure generation (Transforcer module).

Source: Zenodo (CERN European Organization for Nuclear Research)

DOI: 10.5281/zenodo.23038959

Authors: Khokhlov, I. S., Tashchilova, A. S., Seikin, A. A., Yudin, V. S., Keskinov, A. A., Yudin, S. M., Svetlichnyy, D. V., Skvortsova, V. I.

Institutions: State Research Institute of Highly Pure Biopreparations, Federal Medical-Biological Agency

Language: English

Topic: Computational Drug Discovery Methods

Subfield: Drug Discovery

Field: Pharmacology, Toxicology and Pharmaceutics

Domain: Life Sciences

Open Access status: green

OpenAlex ID: W7214944481

Cites:

Cited by:

Related to: 9