DeNoMAP: An Interactive Web Application for Multi-Criteria Prioritization of De Novo Designed Compounds
Year: 2026
Type: preprint
Abstract: De novo molecular generation is increasingly used in drug discovery, yet deciding which generated structures merit follow-up remains difficult, because established benchmarking metrics primarily characterize generative models rather than individual compounds. We present DeNoMAP, a general multi-criteria prioritization framework implemented as an interactive, open-access web application, which positions de novo-generated molecules in a normalized three-dimensional space defined by novelty, synthesizability and drug-likeness. A user-defined desirability zone enables transparent and reproducible candidate selection. Novelty is computed against ZINC20 as an external reference collection using the SmallWorld search engine, rather than against the generative model’s training set. Applied to a published library of 8,066 DNMT1 candidates, the web application identified 15 prioritized compounds among 852 successfully evaluated cluster representatives. DeNoMAP is publicly available at https://denomap.streamlit.app/ and its source code at https://github.com/flfeitosa/DeNoMAP.
Source: ChemRxiv
Language: English
Subfield: Drug Discovery
Domain: Life Sciences
Open Access status: green
OpenAlex ID: W7221813990