Garden has introduced BLOOM (Branching Lookup Optimized for Organic Molecules), a Markush structure search engine designed to provide AI-driven drug discovery teams with near-instant verification of small-molecule intellectual property landscapes. This capability enables researchers to refine molecular candidates with legal certainty integrated directly into the design process. As AI models now generate molecular structures at unprecedented speeds, the primary bottleneck has shifted to due diligence—determining which compounds are already patented before committing laboratory resources and funding. BLOOM is built to eliminate that delay.
The system uses a graph-based, agentic traversal method to evaluate Markush queries against millions of SMILES strings, filtering out invalid candidates early through localized atom and bond feature analysis. It then delivers color-coded mapping to confirm compliance at the atom and bond level, transforming what was once a tedious, manual review into an automated step within the workflow.
Benchmark results show BLOOM achieving a 32.44× average speed increase compared to traditional core-extraction string searches, with times of 0.047 milliseconds per comparison versus 1.491 milliseconds. It also identified accurate matches overlooked by conventional string-based techniques, including successful single-hit queries in datasets containing millions of records. For research teams engaged in IP-conscious design, this means go/no-go decisions can be made during the ideation stage rather than weeks later.
BLOOM significantly reduces false positives, a common limitation in legacy systems that often fail to account for subtle variations in bond counts and placements, sparing scientists from exhaustive manual verification. The tool is fully integrated with Garden’s patent database, linking each SMILES match directly to underlying patent documentation. Garden’s AI agent can then summarize, compare, and streamline result sets to support workflows ranging from rapid novelty assessment to comprehensive freedom-to-operate analysis.
According to Garden’s founder and CEO, Adi Sidapara, BLOOM ensures IP-aware exploration without slowing discovery, while Founding ML Researcher Kavin Sivakumar, Ph.D., notes its ability to capture fine R-group variations at high speed, removing IP checks as a barrier to innovation.
