Allchemy

Allchemy, developed by SciY, is an intelligent chemistry AI platform designed for drug discovery and synthesis planning. It combines state-of-the-art computational synthesis with AI algorithms to predict molecular properties, generate synthesizable lead candidates, and propose efficient synthetic routes — all grounded in expert-curated scientific principles. The platform is built for researchers and chemists who require both accuracy and creative exploration across broad chemical space.

Unlike purely data-driven AI tools that are largely limited to positive examples, Allchemy is founded on expert-curated reaction rules and physical organic heuristics that can describe outcomes not previously reported in the literature. This approach ensures results are reliable, scientifically grounded, and capable of uncovering novel pathways that conventional tools may miss.

Core Capabilities

• Accuracy and Efficiency: Expert-curated rules ensure results are grounded in proven scientific principles, reducing the likelihood of error. Heuristics allow elimination of unfeasible pathways, streamlining synthesis planning.
• Coverage and Innovation: Explores broad chemical space and suggests new, efficient, and sustainable synthetic routes, leading to the discovery of new compounds and materials.
• Continuous Learning: Adapts to new data and evolving scientific knowledge by incorporating new reactions and mechanisms, resulting in continuous improvement of predictions and recommendations.

Platform Features

• More than 40,000 expert-coded reaction rules
• Coverage of more than 600 potentially conflicting functional groups and competing reactions
• Quality of predictions enhanced through incorporation of negative data
• Reaction categorization by scale
• Byproduct tracking throughout synthetic routes
• Evaluation of routes by green chemistry criteria
• AI algorithms for closed-loop synthetic optimization
• Molecular properties evaluation algorithms
• Deployable on individual servers or orchestrated networks

Synthesis Planning Modes

• Retrosynthesis mode: Proposes syntheses for predefined target molecules, ranking pathways for efficiency and greenness, propagating from user-specified substrates, AI-suggested chemicals, or renewable resources.
• Forward synthesis mode: Proposes green chemistry alternatives and provides information about hazardous compounds.
• Within minutes, the platform creates thousands of synthesizable lead candidates meeting user-defined profiles of drug-likeness, affinity towards specific proteins, toxicity, and a range of other physicochemical measures.
• Drug-like scaffolds can be created de novo or evolved from user-defined fragments.

Scientific Validation

• Published in Nature (2022): Computer-designed repurposing of chemical wastes into drugs.
• Published in Science (2022): Closed-loop optimization of general reaction conditions for heteroaryl Suzuki-Miyaura coupling.
• Published in Angewandte Chemie (2024): Machine learning algorithm guiding catalyst choices for magnesium-catalyzed asymmetric reactions.
• Published in Nature (2024): Computational prediction of complex cationic rearrangement outcomes.
• Published in Journal of the American Chemical Society (2024): Demonstrating that AI for retrosynthetic planning needs both data and expert knowledge.

Allchemy supports direct integration with Electronic Lab Notebooks (ELN) and can be deployed on individual servers or as orchestrated networks, making it suitable for a range of organizational scales and workflows in pharmaceutical and chemical research.