Before implementing the CHEESE solution, clients in the early stages of drug discovery faced significant inefficiencies and time losses. Traditional methods for virtual screening and molecular similarity search in large-scale databases (tens of billions of molecules) were slow and impractical. The impact was prolonged development cycles (e.g., saving 200 days of docking score calculations) and difficulties in identifying suitable compounds. Solving this problem was crucial to accelerating early-stage drug development, reducing costs, and gaining a competitive edge.
Acceleration of drug discovery through ultra-fast molecular database search
Client NameInstitute of Organic Chemistry and Biochemistry of the Czech Academy of Sciences
Client CountryCzech Republic, USA
- Client typeOther
- IndustryHealthcare, Pharma & Biotech
- Application areasProduct Development, Innovation, R&D
- AI technologiesAdvanced Analytics / Data Science, Machine Learning
- Business impactsData-Driven Decisions & Planning, Employee Enablement & Productivity
- Data typesOther
- Delivery modelsProduct / Licensed Software, Service / Subscription
- DeploymentsHybrid
- Key capabilitiesOther
- Project stagesScaling / Expanded Implementation
- Solution formsAPI / Micro-service Interface, Web Portal / Dashboard
Solution Description
Problem description
Solution
The developed solution is the CHEESE platform, a set of AI-based tools for drug discovery. Its core function is CHEESE Search, which enables extremely fast 3D similarity searches (shape and electrostatics) in massive chemical databases (tens of billions of molecules) within seconds. The system meets client needs by using AI-generated vectors representing molecules, eliminating the need to generate conformers during search, which significantly increases speed and scalability. With its intuitive user interface and high degree of automation, it enables efficient screening previously unthinkable, accelerating early drug development stages.
Main Users of the Solution
Medicinal chemists, chemoinformaticians
Project timeframe (months)
1
Technologies used
deep learning, latent metric space
Additional services
- Identification and prioritization of suitable use-cases
- Data collection and pre-processing
- AI model selection and customisation
- Systematic AI training programmes
- Ongoing maintenance and retraining of the model
Implementation
Project Owner on the Client's Side
Head of functional/operational unit
Participation on the Client's Side
- Data & ML specialists
- Software & Data Engineering / IT Ops
- End users
Form of Supplier Involvement
Technical support / consultation only
Operation and Maintenance
Operational Model
internal IT customers
Needed Competencies on the Client's Side
Infrastructure/IT team
Other Resources or Infrastructure
Kubernetes/Docker, temporary GPU setup
Impact and Results
Qualitative Benefits
Acceleration of early drug discovery. Expansion of explored candidate molecule space.
Lessons Learned and Recommendations
Key Success Factors
Uniquely solving the problem through close collaboration with domain experts and the supplier’s top AI expertise.
Biggest Challenges
The need to speak the language of biochemistry/drug discovery experts—find an enthusiastic paying partner. This dual motivation is essential.
Recommendation for Others
Try harder 🙂
Promotion
Demo / Public Resources
- CompanyDeep MedChem
- ContactJan Macek
- Emailjan.macek@deepmedchem.com
- Websitehttps://www.deepmedchem.com
- AddressRevoluční 17, 110 00 Praha