DeepMirror

deepmirror is an AI-powered drug discovery platform designed for medicinal chemists and drug discovery teams in both industry and academia. By combining human ingenuity with artificial intelligence, deepmirror empowers teams to focus on the most promising drug candidates during the hit-to-candidate phase, reducing guesswork and accelerating the path to impactful therapeutics. Trusted by over 100 organisations worldwide, the platform integrates broad industry data and AI expertise while keeping proprietary data and intellectual property fully secure.

deepmirror is built to make advanced machine learning accessible to medicinal chemists of all experience levels, acting as an invisible AI partner that enhances creativity and scientific decision-making rather than replacing it.

Core Capabilities

• Generative AI: Explore chemical space at scale to generate novel molecular ideas and candidates that exist beyond standard libraries, revealing high-quality molecules that would otherwise remain hidden.
• Data-driven predictive models: Build high-quality ligand-based models of potency and ADMET properties using proprietary data, enabling teams to predict experimental outcomes with greater accuracy before committing to synthesis.
• Structure-guided design: Apply advanced structure-based design and cofolding algorithms — including models such as AlphaFold3, OpenFold3, Boltz-2, and Chai-1 — to validate design hypotheses, understand binding modes, and refine interactions between ligands and targets.

Key Use Cases and Demonstrated Results

• Screening of 1 billion molecules against cancer targets in just a few hours, enabling the discovery of novel chemotypes against challenging target classes, as demonstrated by Tes Pharma.
• Prediction of experimental biological activity across entire proprietary compound libraries to identify and validate novel active scaffolds.
• Leveraging empirical binding data in conjunction with AI structure prediction models to develop testable design hypotheses for lead compounds, as applied by Curie.Bio.
• Generating actionable solubility predictions to support ADMET profiling workflows.
• Classifying and predicting potency of unknown compounds when applied to large data sets, with generative AI producing chemically sensible suggestions ready for synthesis and testing.

Security and Data Privacy

• Enterprise-grade encryption: All data is encrypted in transit and at rest using modern, highly secure protocols.
• Complete data privacy: User data and models remain strictly private and are never shared without explicit permission.
• ISO 27001 certified: Infrastructure, engineering practices, and policies meet rigorous international security standards, verified by independent audits.

Integrations and Collaborations

• Active collaboration with the Sanders Tri-Institutional Therapeutics Discovery Institute to advance drug discovery research.
• Partnership with Collaborative Drug Discovery to support both humanitarian and commercial drug discovery programmes.
• Collaboration with Medicines for Malaria Venture (MMV) to adapt the machine-learning platform for global health research.
• Post-processing capabilities to make AI-predicted protein–ligand poses chemically interpretable, restoring bond types, hydrogens, and realistic geometry for downstream modelling.

deepmirror is backed by an experienced team of co-founders and advised by industry veterans from organisations including Relay Therapeutics, Bristol Myers Squibb, GSK, and Exscientia. The platform is available via early access and demo request, and is positioned as a democratising force for ML-driven drug design across organisations of all sizes.