Case Study 1: Optimizing Retrosynthesis with SmartChemistry® AI

Example Case Study using ChemAI’s SmartChemistry®

SmartChemistry® successfully identified a cost-effective and scalable synthesis route for Oxaspiroketone (figure 1), a key intermediate in Alzheimer’s clinical candidates. This route has been experimentally validated.

Challenge: Finding an Efficient Route to Oxaspiroketone

A major challenge in synthetic chemistry is designing scalable, high-yield routes for complex molecules. Oxaspiroketone, an important intermediate for Alzheimer’s drug candidates, required an optimized synthesis pathway. Existing AI-driven retrosynthesis models and human-led approaches identified only a five-step synthesis with a low 17% yield – a result constrained by precedent-based reaction similarity.

Figure 1 : Oxaspiroketone Target

In benchmarking the best possible route identified by alternative AI methodologies was a 5 step synthesis as shown in figure 2 with an ultimate 17% yield.  This involved the direct formation of the spiro fragment with all known chemistry available by proximity with examples reported in the literature.

Figure 2 : Direct formation of a spiro fragment

Using its proprietary AI-driven retrosynthesis, SmartChemistry® identified a breakthrough alternative:

Figure 3 : Unique method to generate the spiro fragment identified in SmartChemistry

It is unlikely a chemist would intuitively identify this approach because the disconnection of bond c shown in figure 1 is at odds with the conventional wisdom in chemistry that the disconnection of two atoms both with partial negative charges likely.

✅ A 3-step synthesis route – Reducing complexity and improving efficiency.
✅ Over 40% yield – More than double the yield of traditional methods.
✅ Unconventional bond disconnection – A novel approach not predicted by conventional chemistry wisdom.

Results & Industry Impact

🔹 65% More Viable Routes Identified – Compared to the best alternative AI solutions.
🔹 Validated Experimentally – Demonstrating real-world success in synthesis efficiency.
🔹 Paradigm Shift in Retrosynthesis AI – SmartChemistry® moves beyond precedent-based approaches, unlocking hidden synthetic opportunities.

Conclusion
SmartChemistry® isn’t just an incremental improvement – it redefines what’s possible in retrosynthesis. By leveraging domain-specific AI, curated data, and advanced machine learning, it enables pharma and biotech companies to optimize synthetic routes faster, smarter, and more cost-effectively.

Ready to revolutionize your retrosynthesis? Contact us to learn how SmartChemistry® can accelerate your discoveries.

Case Study 2: Rapid Optimization in Chemistry and Formulation with SmartChemistry® Optimizer

Example Case Study using ChemAI’s SmartChemistry® Optimizer

In industries like pharmaceuticals, cosmetics, agrochemicals, and materials science, formulation and process chemistry are core to performance, but experimentation is often slow, expensive, and inefficient. Traditional Design of Experiments (DoE) methods struggle with:

⚠️ Complex or high-dimensional parameter spaces
⚠️ Learns from noisy, incomplete, or expensive experimental data
⚠️ Limited support for categorical variables
⚠️ Slow iteration and low scalability

Chemists are left to rely on trial-and-error, wasting time and resources.

Solution: ChemAI developed the SmartChemistry® Optimizer – an AI-driven optimization platform co-designed with computer scientists and powered by advanced algorithms tailored to chemical R&D.

Unlike conventional methods, SmartChemistry®:

✅ Requires no molecular formula input
✅ Learns from noisy, incomplete, or expensive experimental data
✅ Supports categorical variables and high-dimensional space
✅ Is intuitive for chemists to use, with no coding or data science skills required

Process Chemistry – Pharma Case Study

Scope: 11 projects over 12 months

Impact:
▪ Outperformed traditional methods in 6 out of 11 campaigns
▪ Achieved breakthrough efficiency in 2 campaigns (e.g. 50 experiments reduced to just 15)
▪ Delivered equal performance in the remaining 3 campaigns
▪ Detected flawed data early, avoiding wasted experiments
▪ Cut experiment counts by over two-thirds in top-performing scenarios

Formulation Science – Cosmetics Case Study

Objective: Optimize a bio-based skin cream formula with no prior data

Impact:
▪ Just 8 AI-driven iterations
▪ Only 40 total experiments
▪ Delivered a formulation considered impossible using traditional approaches

“Very difficult to design this formulation without SmartChemistry AI.”
– Lead Data Scientist, Cosmetics Company

Why It Matters

SmartChemistry® Optimizer is built for real-world chemistry and formulation, where data is imperfect and stakes are high. It learns, adapts, and optimizes – faster and smarter than human-guided experimentation alone.

For any organisation aiming to reduce cost, accelerate discovery, and increase confidence in their chemistry R&D – SmartChemistry® Optimizer delivers measurable results.

Ready to optimize smarter, faster, and with fewer experiments? Contact us to see how SmartChemistry® can transform your R&D efficiency.

Case Study 3: Transforming ELN Data and CRO Reports into High-Quality Machine Learning Datasets with SmartChemistry® Curation

Challenge

R&D organizations accumulate massive volumes of valuable chemical data from Electronic Lab Notebooks (ELNs), Contract Research Organizations (CROs), and Contract Development & Manufacturing Organizations (CDMOs). But much of this data remains trapped in free-text formats — unstructured, inconsistent, and unusable for machine learning or automation.
Common challenges include:

🔸 Inconsistent formats and chemical notations
🔸 Limited accessibility due to unstructured, free-text reports
🔸 Sparse, heterogeneous data across sources
🔸 Missing metadata and experimental context

Solution

SmartChemistry® Curation is an AI-powered platform designed specifically to extract, structure, and enrich unstructured chemistry data at scale. Unlike generic NLP tools, it integrates domain-specific cheminformatics with advanced language models to deliver structured datasets optimized for ML, AI, and automation workflows.

Key capabilities:

▸ Breaks down complex reaction procedures into atomic steps
▸ Extracts key experimental details — chemicals, parameters, conditions, equipment
▸ Parses ChemDraw® schemes and resolves chemical structures
▸ Standardizes units, names, and notations using trusted chemical databases
▸ Outputs structured formats (XML, JSON, databases) ready for downstream applications
▸ Achieves over 90% curation accuracy, processing recipes in under 3 minutes

Results

Case Study – Structuring CRO & ELN Data at Scale

Context: A leading chemical organization needed to unlock insights from decades of CRO and ELN reports.

Traditional NLP tools failed to extract key steps and chemical entities from typical recipes.
With SmartChemistry® Curation:

▸ All chemicals and molecular representations were correctly identified
▸ Eight procedural steps were classified into a chemical operations ontology
▸ Equipment, parameters, and conditions were fully extracted
▸ A 750-line structured XML dataset was generated – instantly usable for ML, AI, and automation

This level of performance is unmatched by traditional tools and has led multiple global R&D organizations to adopt SmartChemistry® Curation as a core part of their data infrastructure.

SmartChemistry® Curation transforms unstructured chemical knowledge into AI-ready datasets – enabling advanced discovery, faster automation, and new insights from historical experiments.
It’s not just data extraction – it’s data transformation for the future of chemistry.

Ready to unlock the full value of your chemical data? Contact us to see how SmartChemistry® Curation can prepare your R&D for the AI era.

Case Study 4: Yield Prediction Models for Medicinal and Process Chemistry with SmartChemistry® Generation

Sector: Medicinal & Process Chemistry
Transformation Focus: Amide Bond Formation
Technology Stack: Machine Learning, Robotic Automation, Active Learning, UPLC-MS

Challenge

A global CRO needed to reduce the number of physical experiments required to optimize reaction conditions for amide bond formation -the most frequently used transformation in their workflows.

Traditional methods were:

🔸time-consuming
🔸resource-intensive
🔸limiting throughput & innovation speed.

Solution

ChemAI deployed its SmartChemistry® platform to build a predictive model capable of forecasting reaction outcomes using machine learning. By integrating AI-guided experimental design with robotic synthesis and automated LC-MS analysis, the system significantly reduced experimental load and improved predictability.

Results

✅50% reduction in required experiments through active learning
✅Fully automated LC-MS data processing pipeline
✅Predictive model generalizable to new reactants
✅Morgan fingerprints identified as optimal descriptors
✅Achieved F1-score of 0.68 (Gold validation level)
✅Plate 2 designed using Gaussian Process modelling for uncertainty-based selection

With expanded data and refined modeling strategies underway, the project is now progressing toward ternary classification and quantitative yield prediction with a target accuracy of 80–90% – positioning the system for routine deployment in both medicinal and process chemistry environments.

Case Study 5: Accelerating Polymer Formulation Development with AI using SmartChemistry® Optimizer

Sector: Polymer R&D
Use Case: Formulation Optimization
Technology Stack: SmartChemistry® Prediction & Optimization Modules, Machine Learning, Experimental Design Automation

Challenge

A leading polymer compound developer needed to speed up formulation development across diverse sectors including automotive, electronics, medical devices, and more. Each new compound involved 15+ formulation and process parameters, requiring up to 100 trial-and-error experiments per project. Under growing pressure to shorten R&D cycles without compromising product quality, they sought a smarter, more scalable solution.

Solution

Through a six-month Proof of Concept, ChemAI implemented its SmartChemistry® platform to guide formulation design with data-driven intelligence. We worked closely with formulation scientists across five R&D projects, providing:

▸AI-powered formulation design using Prediction and Optimization modules
▸Guidance on structuring and importing historical data
▸Extensive training in AI best practices and experimental design
▸Iterative support for building and refining experimental campaigns

SmartChemistry® proved effective even with limited initial data – supporting both data-rich and low-data projects.

Outcome

✅40–60 fewer experiments per project
✅€600,000 estimated annual R&D cost savings
✅Effective modeling even with <10 starting experiments
✅Scalable implementation potential across all formulation labs
✅Ongoing improvement in prediction accuracy as more data is integrate

With SmartChemistry® now embedded in their workflow, this partner is set to continuously improve formulation productivity, reduce development timelines, and unlock greater innovation across every R&D initiative.