Timothy R. Newhouse

Herein, we report a practical cobalt-catalyzed method for the ( E )-stereoselective synthesis of enamides and N -alkenyl heterocycles via positional isomerization of N -allyl amides and nitrogen heterocycles.

This work describes the development of a de novo strategy to access the A,D-seco limonoids utilizing robust synthetic operations allowing for the synthesis of a characteristic member, deoxylimonin. The retrosynthetic strategy uses resilient chemistry for the assemblage of a key carbocyclic tricycle in only 8 synthetic maneuvers. This intermediate is then amalgamated through a modern variation on the classic combination of ring cleavage and formation using an oxy-Michael reaction as a key synthetic operation. Two discrete local computational models were developed and rapidly integrated into the design stages of retrosynthetic analysis, illustrating how multiple local models augment the intersection of transform-based and structure-goal-based approaches.

The exponential growth of scientific literature presents an increasingly acute challenge across disciplines. Hundreds of thousands of new chemical reactions are reported annually, yet translating them into actionable experiments becomes an obstacle^1,2. Recent applications of large language models (LLMs) have shown promise^3,4,5,6, but systems that reliably work for diverse transformations across de novo compounds have remained elusive. Here we introduce MOSAIC (Multiple Optimized Specialists for AI-assisted Chemical Prediction), a computational framework that enables chemists to harness the collective knowledge of millions of reaction protocols. MOSAIC is built upon the Llama-3.1-8B-instruct architecture⁷, training 2,498 specialized chemical experts within Voronoi-clustered spaces. This approach delivers reproducible and executable experimental protocols with confidence metrics for complex syntheses. With an overall 71% success rate, experimental validation demonstrates the realizations of over 35 novel compounds, spanning pharmaceuticals, materials, agrochemicals, and cosmetics. Notably, MOSAIC also enables the discovery of new reaction methodologies that are absent from the expert’s training, a cornerstone for advancing chemical synthesis. This scalable paradigm of partitioning vast domains into searchable expert regions enables a generalizable strategy for AI-assisted discovery wherever accelerating information growth outpaces efficient knowledge access and application.

This work describes the development of a de novo strategy to access the A,D-seco limonoids utilizing robust synthetic operations allowing for the synthesis of a characteristic member, deoxylimonin. The retrosynthetic strategy uses resilient chemis-try for the assemblage of a key carbocyclic tricycle in only 8 synthetic maneuvers. This intermediate is then amalgamated through a modern variation on the classic combination of ring cleavage and formation using radical attack at oxygen in a key synthetic operation. Two unique local computational models were developed and rapidly integrated into the design stages of retrosynthetic analysis illustrating how multiple local models augment the intersection of transform-based and structure-goal based approaches.

The selection of synthetic routes to a small molecule of interest is enabled by the use of various tools to assess the chemical complexity of a given intermediate. While prior approaches assess the intrinsic molecular complexity or the facility with which an intermediate can be synthesized, in this study, we introduce an alternative approach that tracks the progress toward the target structure in a given synthesis. A simple metric, EvolvedComplexity, was developed that compares the chemical similarity of a pair of molecules on the basis of the Tanimoto distance between chemical fingerprints. This complementary approach to assessing progress in the synthesis may prove to be a useful tool for planning synthetic routes and developing novel chemistries.

Herein, we report the total synthesis of erythrocarpines M and L and thaigranatin P, topologically complex phragmalin-type limonoids, the latter exhibiting potent agonism on the human pregnane X receptor (hPXR). A key transformation is a SmI₂-mediated reductive coupling of an aldehyde and an alkene, which constructs the signature tricyclo[3.3.1^2,10.1^1,4]decane framework characteristic of phragmalin natural products. The synthesis features a strategically convergent construction of the phragmalin core structure via a cuprate addition, triflation, and intramolecular Heck reaction sequence. This route offers a versatile platform for the synthesis of structurally diverse phragmalin analogs for future biological investigation.

Azobenzenes are proven to be one of the most successful molecular photoswitches applied across different fields such as organic chemistry, material science, cosmetics, and pharmaceuticals. Such a widespread implementation is possible because of their photochromic properties contingent upon the substitution pattern and aryl-core nature. In recent endeavors of molecular design, replacing one or both phenyl rings with heteroaromatics turned out to be a good strategy to access compounds with improved photoswitching properties, as well as to expand molecular diversity. One of the challenges related to the design of new azobenzene photoswitches is that it often includes the synthesis of large libraries of compounds due to limited methods for prediction of their properties. Herein, we present a computationally driven workflow for the design and synthesis of a novel class of azobenzene photoswitches, heteroaryl azobenzenes with N-bridgehead heterocycles─pyrazolo[1,5-a]pyrimidine and 1,2,4-triazolo[1,5-a]pyrimidine. A small library of heteroaryl photoswitches was synthesized, and their photochemical properties were evaluated. Subsequently, these results were used to validate a computational approach, which included the in silico evaluation of a large library of designed photoswitch candidates leading to the synthesis of a new photoswitch with improved spectral properties, red-shifted λ_max values.

While hundreds of thousands of new chemical reactions are reported annually, efficient use of this vast collection of synthetic knowledge remains a persistent challenge in modern chemistry. Recent applications of large language models (LLMs) have shown promise, but systems that reliably work for de novo compounds and molecular transformations have remained elusive. Here we introduce MOSAIC (Multiple Optimized Specialists for AI-Driven Chemical Prediction), a computational framework that enables chemists to harness the collective knowledge of millions of reaction protocols. In contrast to existing approaches relying on agentic models, MOSAIC leverages the open-source Llama3.1-8B-instruct architecture. By training 2,489 specialized chemical experts on Voronoi-clustered reaction spaces, we establish a scalable paradigm that delivers reproducible and human-readable experimental protocols for complex syntheses. Experimental validation demonstrates MOSAIC's ability to predict and execute previously unreported transformations, including challenging reactions via Buchwald-Hartwig amination, Suzuki coupling, and olefin metathesis. We validate this approach through the successful synthesis of over 35 novel compounds spanning pharmaceuticals, materials, agrochemicals, and cosmetics. This framework establishes a new relationship between computational and experimental chemistry, providing a foundation for accelerated chemical discovery across disciplines.

The selection of synthetic routes to a small molecule of interest is enabled by the use of various tools to assess the chemical complexity of a given intermediate. While prior approaches assess the intrinsic molecular complexity or the facility with which an intermediate can be synthesized, in this study we introduce an alternative approach that tracks the progress towards the target structure in a given synthesis. A simple metric, EvolvedComplexity, was developed that compares the chemical similarity of a pair of molecules on the basis of the Tanimoto distance between chemical fingerprints. This complementary approach to assessing progress in synthesis may prove to be a useful tool for planning synthetic routes and for developing novel chemistries.

Deconstructive hydrodealkenylations are valuable transformations, particularly utilized in the chiral pool-based enantiospecific syntheses. Current approaches to hydrodealkenylations have limitations regarding functional group tolerance, with ozone being a primary reagent in the installation of the peroxyketal intermediate formed prior to Fe(II) induced β-scission. This work details an ozone-free protocol with a selectivity profile that is expected to follow the Mukaiyama peroxidation.