Resource-aware AI for drug discovery

Technology

Allchemy’s drug-discovery platform combines state-of-the-art computational synthesis with AI algorithms to predict molecular properties. Within minutes, Allchemy creates thousands of synthesizable lead candidates meeting user-defined profiles of drug-likeness, affinity towards specific proteins, toxicity, and a range of other physical-chemical measures. At the push of a button, drug-like scaffolds are created de novo or evolved from user-defined fragments; syntheses are ranked for efficiency and greenness, and are propagated from either user-specified substrates, AI-suggested chemicals, or renewable resources.

Some of our tools are freely available for you to try. For example, the use of Allchemy’s “Life” module with which to explore prebiotic reactions is available at LINK (short youtube summary is here for detailed user manual see SI to our Science paper). Another freeware illustrates the performance of our AI algorithms to predict molecular properties – at this link, you can play with the tool to predict pKa’s of C-H acids (for details, see JACS). These and other modules are integrated into Allchemy’s professional WebApp – to learn more, please arrange a demo at .

Allchemy encompasses the entire resource-to-drug design process and has been used in academic, corporate and classified environments worldwide to:

Design synthesizable leads targeting specific proteins
Evolve scaffolds similar to desired drugs
Design “circular” drug syntheses from renewable materials
Interface with and instruct automated synthesis platforms and optimize pilot-scale processes
Operate “iterative synthesis” schemes
Predict side reactions and create forensic “synthetic signatures” of hazardous/toxic molecules
Design synthetic degradation and recovery cycles for various types of feedstocks and functional target molecules

People

We are a team of synthetic and medicinal chemists and computer-specialists at the forefront of modern “chemical AI” revolution. Some of us has previously developed the Chematica retrosynthesis platform (cf. recent Nature paper). We have been working closely with both the private sector as well as US Government agencies (especially DARPA).

Agnieszka Wołos

Senior Chemist, in silico drug discovery

Karol Molga

Senior Chemist, computational reaction design

Wiktor Beker, PhD

Senior AI Specialist

Rafał Roszak, PhD

Senior Software-Development Specialist

Tomasz Klucznik, PhD

Senior Chemist, computational reaction design

Barbara Mikulak-Klucznik, PhD

Synthetic Chemist, computational reaction design

Martyna Moskal

Software Engineer

Sara Szymkuć, PhD

Co-Founder and President

News

December 2020

Our President, Dr. Szymkuc interviewed by Nature and National Geographic!

October 2020

Allchemy commences Phase II of DARPA’s Accelerated Molecular Discovery, AMD, program. We are working as part of the MADNESS team – why “MADNESS”? Well, check out our partners in crime: Aspuru-Guzik, Cronin, Hein and Burke’s labs! Our algorithms and their robots are hell of a madness mix!

September 2020

Our Science paper is out!  See how Allchemy has been used to discover new, lab-validated prebiotic syntheses and self-regenerating cycles. As a bonus, feel free to play with Allchemy’s “Life” module available here.

August 2020

We are proud to become an Affiliate of the Molecule Maker Lab Institute, one of only five AI Centers funded by the NSF. We will be working with colleagues at Urbana-Champaign to develop new robotized chemistries and also new catalysts. Fun science ahead!

August 2020

Our AI is spot on when it comes to distinguishing drug-like from non-drug-like molecules! Read our newest paper in Nature Machine Intelligence

October 2019

One of many properties that Allchemy considers during synthesis planning is pKa of C-H acids. Read our first JACS paper describing how these pKa’s can be predicted with unprecedented precision. The software is freely available here.

August 2019

Work is underway! We are excited to announce that Allchemy has commenced effort on two DARPA-sponsored projects, one with Lawrence Livermore National Laboratory and one within the Accelerated Molecular Discovery program.

Publications

Science

Synthetic connectivity, emergence, and self-regeneration in the network of prebiotic chemistry

A. Wołos, R. Roszak, A. Żądło-Dobrowolska, W. Beker, B. Mikulak-Klucznik, G. Spólnik, M. Dygas, S. Szymkuć, B. A. Grzybowski

Nature Machine Intelligence

Minimal-uncertainty prediction of general drug-likeness based on Bayesian neural networks

W. Beker, A. Wołos, S. Szymkuć, B. A. Grzybowski

Journal of American Chemical Society

Rapid and accurate prediction of pKa values of C–H acids using Graph Convolutional Neural Networks

R. Roszak, W. Beker, K. Molga, B. A. Grzybowski

Science

Synthetic connectivity, emergence, and self-regeneration in the network of prebiotic chemistry

A. Wołos, R. Roszak, A. Żądło-Dobrowolska, W. Beker, B. Mikulak-Klucznik, G. Spólnik, M. Dygas, S. Szymkuć, B. A. Grzybowski

Nature Machine Intelligence

Minimal-uncertainty prediction of general drug-likeness based on Bayesian neural networks

W. Beker, A. Wołos, S. Szymkuć, B. A. Grzybowski

Journal of American Chemical Society

Rapid and accurate prediction of pKa values of C–H acids using Graph Convolutional Neural Networks

R. Roszak, W. Beker, K. Molga, B. A. Grzybowski

Contact Us