The Science Behind Denovo Platform

Step 1

Targeting The Best Conformation

Accurate structure-based drug design begins with selecting the right protein conformation, a factor we have previously demonstrated to be critical for maximizing the performance of computational screening and molecular design campaigns. We have developed Proteus, a proprietary platform that systematically analyzes and compares all available conformations of a given target deposited in the Protein Data Bank.

It performs automated benchmarking of each conformation using known active compounds and carefully designed decoys, quantitatively evaluating their ability to distinguish true binders from non-binders. Based on these objective performance metrics, the platform recommends the optimal protein conformation for structure-based drug design and virtual screening, establishing a reliable foundation for downstream discovery efforts. Proteus is available both as a standalone product and as an integrated module within the Denovo Platform, offering flexible deployment to meet the needs of different discovery workflows.

Q-pocket is available as a standalone product and as an integrated module within the Denovo Platform

Step 2

Designing Hit Candidates

Unlike traditional virtual screening, which is constrained to evaluating pre-existing chemical libraries, the Denovo Platform employs proprietary generative algorithms that design small molecules de novo directly within the target binding pocket. Our models explicitly account for the three-dimensional shape, physicochemical properties, and interaction patterns of the binding site, constructing candidate molecules atom by atom to maximize complementarity with the target.

This approach overcomes the limitations of virtual screening, including library bias, limited chemical diversity, and poor coverage of relevant chemical space. Rather than searching for the best “available” molecule, Denovo generates molecules that are purpose-built for the target from the outset. As a result, our approach delivers higher-quality hit candidates with improved binding potential, drug-like properties, and optimization readiness, while dramatically reducing the number of compounds that need to be explored downstream.

The Denovo generative design engine is provided as an integrated module within the Denovo Platform for structure-guided hit discovery.

Step 3

Selecting Superior Candidates

The critical step in early drug discovery is the selection of the right hit candidates for laboratory experiments. At this stage, we focus on identifying molecules that not only exhibit strong target engagement potential but also possess the characteristics required to become real, developable drug candidates.

A key component of this process is the assessment of the synthetic feasibility of compounds developed by our generative AI models. To address it, we developed Synthony, a proprietary AI-driven tool specifically designed to predict the synthesizability of highly novel compounds with exceptional speed and accuracy.

Synthony evaluates plausible synthetic routes, chemical complexity, and reaction accessibility, enabling rapid prioritization of molecules that can be synthesized efficiently in real-world laboratory settings. This extra-fast synthesis prediction allows us to eliminate impractical designs early, significantly reducing time and cost in medicinal chemistry campaigns.

In parallel, we evaluate the drug-likeness of each compound using our holistic, AI-based drug estimation system. Unlike traditional rule-based filters, this system integrates more than 100 distinct features spanning physicochemical properties, structural complexity, stability, developability, and optimization potential. By combining these diverse signals into a unified framework, the platform provides a nuanced and actionable assessment of each molecule’s likelihood of progressing into a viable drug candidate.

Synthony and Denovo’s drug-likeness assessment tools are integrated modules within the Denovo Platform for early candidate selection.