From Ghana to Berkeley
Sarpong's path took him from Macalester College to Princeton, where he worked with Martin Semmelhack (another Molekula advisor), then to Caltech and Berkeley. In 2025, he was elected to the National Academy of Sciences. He also won the Inhoffen Medal, one of Germany's top prizes for natural products chemistry.
His lab synthesizes complex molecules from nature, the kind that might become drugs. Over 180 papers so far.
Computers Planning Syntheses
Here's why Sarpong matters to Molekula: he runs one of the main research groups figuring out how computers can help plan chemical syntheses.
He's a principal investigator for C-CAS, the NSF Center for Computer-Assisted Synthesis. The goal is to combine machine learning with chemistry to change how chemists make complex molecules.
Traditional synthesis planning works like this: a chemist looks at a target molecule, thinks about what reactions might build it, checks their memory for similar examples, and sketches out a plan. This takes years to learn and depends entirely on what you happen to remember or can find.
Computer-assisted synthesis flips this. The system analyzes chemical literature databases, finds relevant examples, and proposes routes. Things that might take a chemist days to figure out can happen in minutes.
Testing Synthia
In 2021, Sarpong tested a program called Synthia on pupukeanane natural products (complex molecules from marine sources). The results showed both what works and what doesn't.
Synthia generated synthesis proposals in minutes. It found routes using established chemistry and included an integrated Reaxys search to find similar examples from literature.
But the best results came when chemists used Synthia proposals as starting points, then applied their own expertise to improve them. The computer wasn't replacing the chemist. It was giving the chemist better tools.
This matters. Computer tools work best when they augment human expertise, not try to replace it.
The Real Bottleneck
Here's the connection to Molekula: Synthia's integrated Reaxys search was essential. Even cutting-edge AI synthesis planning needs fast access to chemical literature.
But Reaxys and SciFinder are expensive, slow to index new research, and hard to use. Sarpong's work showed that Synthia "greatly accelerated synthesis planning" by quick access to literature examples, but this speed hits a wall when the databases themselves are slow.
Think about what computer-assisted synthesis could do with literature that was:
- Indexed in hours, not months
- Searchable in plain English
- Comprehensive across journals, preprints, patents
- Affordable for any researcher
That's the infrastructure needed to make tools like Synthia work at scale.
Beyond Just Planning
Sarpong's recent work goes past just planning syntheses. His 2021 primer in Nature Methods Primers (with the Doyle and Cernak labs) looked at both planning and executing automated synthesis.
The vision: computers plan the synthesis, robots execute it. But every step needs literature knowledge. Which reactions work reliably? What conditions? Which protecting groups stay stable? All of this lives in chemistry papers.
His 2024 paper in Nature Reviews Chemistry on molecular complexity explored how analyzing complexity can guide strategy. Again, this needs comprehensive literature access to see what's been done and what worked.
Teaching Matters
Sarpong won the UC Berkeley Chemistry teaching award three times (2009, 2024, 2025) and the 2016 Noyce Prize for undergraduate teaching.
When you're training new chemists, literature access isn't optional. Grad students shouldn't spend weeks learning database syntax before starting research. Undergrads shouldn't be locked out of recent papers because their school can't afford subscriptions.
Why This Matters
Sarpong's work shows the future of chemistry is hybrid: human creativity plus computational tools. But making that work requires solving literature access.
Computer synthesis planning is only as good as the literature it can search. AI retrosynthesis tools need comprehensive reaction databases. Machine learning models predicting reactions need training data from published papers.
The bottleneck isn't computing power. It's knowledge access.
This is why Molekula AI's mission resonates with the work happening in labs like Sarpong's. C-CAS aims to create tools for a new generation of "data chemists" who can leverage computational methods for molecular synthesis. But data chemists need data, and that data lives in the chemistry literature.
What Needs to Happen
Sarpong noted that computer tools help most "for early-stage career scientists as well as those focused on identifying synthesis routes backed by precedent." The researchers who need better literature access most (grad students, postdocs, scientists at smaller schools) face the highest barriers right now.
Work from Berkeley and C-CAS shows what's possible when computers meet chemistry expertise. But it needs infrastructure: fast, comprehensive, affordable access to chemistry knowledge.
That's not just technical. It's required for the future Sarpong and colleagues are building, where computers help chemists work faster and knowledge access is open to everyone.
