Garden Launches BLOOM to Make AI Drug Design IP-Aware in Real Time

Garden has introduced BLOOM (Branching Lookup Optimized for Organic Molecules), a Markush structure search engine designed to give AI drug-design teams near-instant insight into small-molecule IP landscapes, enabling researchers to iterate on candidates with built-in legal certainty. As AI models generate molecules at unprecedented speed, the main bottleneck has shifted to diligence—determining what’s already patented before investing lab time and resources. BLOOM removes that delay.

The engine employs a graph-based, agentic traversal to match Markush queries against millions of SMILES strings, rapidly discarding invalid candidates using local atom and bond features. Results are presented with color-coded mapping that confirms atom- and bond-level compliance, turning IP verification from a manual, error-prone process into a seamless design-stage step.

In benchmark testing, BLOOM achieved an average 32.44× speed boost over standard core-extraction string searches (0.047 ms vs. 1.491 ms per comparison) while identifying correct matches missed by string-based methods, including a single-hit query across a multi-million-record corpus. This enables go/no-go decisions during ideation rather than weeks later, and reduces false positives that often result from legacy tools’ difficulty handling nuanced bond counts and positions—saving teams from tedious atom-by-atom reviews.

BLOOM is integrated with Garden’s patent database, linking every SMILES match to the corresponding patent records. Garden’s AI agent can then summarize, compare, and help refine result sets, supporting workflows from rapid novelty screening to freedom-to-operate analysis alongside AI-driven molecular design.

“AI can propose thousands of viable chemistries in minutes. BLOOM closes the loop by telling you what’s already fenced off instantly,” said Adi Sidapara, founder and CEO of Garden. “You get IP-aware exploration without slowing down discovery.” Founding ML Researcher Kavin Sivakumar, Ph.D., added, “Small changes around an R-group can define patentability. BLOOM’s graph reasoning captures those subtleties at speed, so IP checks no longer throttle design.”