MLConfGen
Spatially Aware Molecule Generation: Faster Discovery, Greater Efficiency, Higher Hit Rates
Small-molecule generation for real-world workflows
Find on Marketplaces →Key Capabilities & Value
Spatially aware generation
Physics-grounded spatial conditioning without expensive descriptors or complex interaction profiling. Spatial intent built in from the start
Lightweight by design
Under 200 MB total model weight. Inference memory stays below ~14 GB even for batches of 100 samples. Designed to run on modest hardware — not just on high-end clusters
Flexible deployment
SaaS, on-prem, or cloud deployment: flexible options for any infrastructure
Fixed fragment generation
Native fixed-fragment generation with scaffolds anchored in place during generation
Pipeline-ready architecture
MLConfGen is designed as a modular proposal engine, not a monolithic end-to-end system. Pairs naturally with geometry optimization, docking, and free-energy calculations without displacing your existing tools
Synthesis-aware outputs
Synthesis-ready molecule generation with physically plausible, low-strain geometries for downstream use
Where MLConfGen Lives in Your Pipeline
MLConfGen doesn't try to replace your downstream tools. It makes them more efficient by generating a targeted, spatially constrained conformer pool that docks better, filters faster, and costs less to process at every subsequent stage.
Related News
Quantori MLConfGen Wins “Machine Learning Innovation Award” in 2026 AI Breakthrough Awards Program
Ninth annual program honors the AI innovators shaping the next era of global innovation.
Recording available: Run GPU molecular simulations webinar
Watch the recap where we demonstrate an end-to-end automated molecule generation workflow: from pocket detection to spatially aware generative ligand design.
Spatial Conditioning for Small-Molecule Generation: Expressivity, Flexibility, and Deployment
By Denis Sapegin, PhD, Quantori Principal Cheminformatics Engineer