Connor Coley
Massachusetts Institute of Technology, MA·Chemistry·Early-career
Making new molecules for medicines, polymers, agrochemicals, catalysts and organic electronics depends on organic synthesis, the chemistry of building carbon-based molecules step by step. Planning and running these reactions still relies heavily on chemists' intuition and trial and error.
- Fields
Chemistry
Computer Science
- Topics
machine learning
organic synthesis
synthetic organic chemistry
- Grants
- 1 known grant, through Feb 2027
What they work on
- Test whether reaction condition optimization becomes more efficient when algorithms use representations learned from the chemical literature as priors.
- Build reaction-prediction models grounded in physical organic chemistry principles and test whether they generalize better to new reactions than domain-agnostic machine learning models.
- Develop novelty detection algorithms that characterize significant discoveries in organic synthesis, and use them to bias de novo generation of new reactions.
- Resolve knowledge gaps such as uncertain substrate compatibility using query-driven active learning and Bayesian optimal experimental design, which pick the most informative experiments to run.
More about the research
Machine learning, where computers learn patterns from data, has started to help with isolated tasks such as planning synthesis routes and tuning reaction conditions. However, current models need huge amounts of data, generalize poorly to new chemistry, ignore established chemical theory and have not yet made real discoveries. This project aims to build machine learning tools tailored to chemistry that combine experimental data, theoretical chemistry knowledge and modeling, so that synthesis becomes more predictable and robust.
Why it matters
Faster and more predictable synthesis could lower the cost and time of making new chemical compounds. This would help chemists working on new medicines and on sustainability technologies such as catalysts and organic electronics. Open-source tools from the project could also give synthetic chemists data-driven support for discovering and developing new reactions.
Known grants
From NIH and NSF research grants. Labs often have other funding we can't see.
Closest labs
Dasan Thamattoor
Colby College, ME·Chemistry
Generate singlet vinylcarbenes from cyclopropanated phenanthrene precursors and determine their structure and reactivity, given their proposed role in cyclopropene isomerizations and related synthetic transformations.
Timothy Cernak
University of Michigan, MI·Chemistry·Early-career
Use graph theory, including adjacency matrix mapping of known total syntheses of terpenes, alkaloids and pharmaceuticals, to identify reaction types that do not yet exist but would be valuable for synthesis.
Song Lin
Cornell University, NY·Chemistry
Develop hypervalent iodine reagents carrying labile ligands that selectively couple two unactivated aliphatic amines to form N–N bonds.
The school
Massachusetts Institute of Technology: 25 labs on PI Pages, $1.1B a year on research, 692 research doctorates a year.
Reach out
A start for your email. Fill in the brackets.