Automated synthesis planning has become an important tool in drug discovery.One exciting application is de novo drug discovery, in which automated synthesis planning can be used to prioritize synthetically accessible mols.However, it can take minutes to plan synthesis for a drug-like mol. on a single CPU.While relatively fast, this can be too slow for applications such as de novo drug discovery.To address this challenge, recent work has focused on speeding up automated synthesis planning software.Bringing down processing time would also more broadly improve user experience of retrosynthesis software.Two approaches have recently emerged.The first idea is to precompute outputs of synthesis planning software for a large number of mols.We can then train a neural network on such a dataset to rapidly predict outcomes of retrosynthesis software for novel compoundsThe second approach is to use reinforcement learning techniques to learn from previously executed plans to prioritize successful choices in future searches, as done by Kim et al. and Schreck et al.In this talk, we will discuss these and other techniques for rapid estimation of synthetic accessibility.To compare them on even grounds, we trained two recently introduced techniques on the same dataset, and evaluated them against a comprehensive set of benchmarks.

Predicting reaction outcome using ML models trained on dedicated datasets of reactions performed in HTE mode

Author: van Workum, Ruard ; Rzymkowski, Jan ; Ulatowski, Filip ; Wach, Paulina ; Choluj, Artur ; Jastrzebski, Stanislaw ; Byrski, Piotr ; Wlodarczyk-Pruszynski, Pawel ; Szczupak, Lukasz

In modern organic chem. dealing with more and more complex structures, the success of any chem. reaction is highly unpredictable.It often comes to broad reaction condition optimization before the expected product is obtained.Machine learning (ML) models become more useful in many aspects of organic chem. such as mol. modeling, predicting properties (i.e. solubility, toxicity), retrosynthesis, or synthetic availability scoring.ML models capable of predicting reaction outcome would be extremely useful in reducing the number of failed reaction or in selecting building blocks used in combinatorial chem.Application of ML models for predicting reactivity for specific reactants is very challenging mainly due to imperfections of available datasets of chem. reactions, which are strongly biased toward successful reactions, and exhibit no standardization of reaction conditions.In order to tackle this challenge we generated own dataset of chem. reactions consisting of subsets of approx. 10K reactions performed under identical reaction conditions with a large variety of reactants.Models trained on this dataset achieved ROC AUC values >70% in predicting the success of chem. reactions, even for complex drug-like reactants not present in the training data.Remarkably, in this task, ML models outperformed a representative group of organic chemists with PhD-level expertise.These promising results show the potential of ML models combined with properly structured exptl. datasets as a transformative tool in addressing the challenges posed by the evolving landscape of complex organic chem.

Molecule-Edit Graph Attention Network: Modeling retrosynthesis prediction as a sequence of edits

Author: Jastrzebski, Stanislaw ; Wlodarczyk-Pruszynski, Pawel ; Sacha, Mikolaj ; Byrski, Piotr ; Blaz, Mikolaj

Retrosynthesis - the task of finding a set of reactants that can be used to synthesize a target mol., is one of the fundamental problems in organic chem.Recently, there have been substantial developments in the field of computer-aided retrosynthesis.Most of such approaches employ reaction templates, which are able to give highly accurate and interpretable predictions but suffer from low generalization due to the limited coverage.On the other hand, template-free methods that view retrosynthesis simply as a translation problem lack interpretability and are prone to making trivial mistakes.In this work, we present the Mol.-Edit Graph Attention Network (MEGAN) - a template-free model that encodes reactions as series of operations on mol. graphs.Our model is highly scalable and does not require any manual rule encoding.Moreover, the sequence of edits generated by MEGAN provides mapping of atoms between product and substrates.MEGAN surpasses all published template-free and template-based models on a popular USPTO-50k retrosynthesis benchmark in terms of top-10 accuracy while achieving above 95% coverage of the held-out validation reaction dataset.

100 Deals associated with Molecule One Sp zoo

100 Translational Medicine associated with Molecule One Sp zoo

Corporation Tree

Boost your research with our corporation tree data.

view full example data

Pipeline

Pipeline Snapshot as of 12 Mar 2026

No data posted

Deal

Boost your decision using our deal data.

view full example data

Translational Medicine

Boost your research with our translational medicine data.

view full example data

Profit

Explore the financial positions of over 360K organizations with Synapse.

view full example data

Grant & Funding(NIH)

Access more than 2 million grant and funding information to elevate your research journey.

view full example data

Investment

Gain insights on the latest company investments from start-ups to established corporations.

view full example data

Financing

Unearth financing trends to validate and advance investment opportunities.

view full example data

AI Agents Built for Biopharma Breakthroughs

Accelerate discovery. Empower decisions. Transform outcomes.

Get started for free today!

Accelerate Strategic R&D decision making with Synapse, PatSnap’s AI-powered Connected Innovation Intelligence Platform Built for Life Sciences Professionals.

Start your data trial now!

Synapse data is also accessible to external entities via APIs or data packages. Empower better decisions with the latest in pharmaceutical intelligence.

Bio

Bio Sequences Search & Analysis

Chemical

Chemical Structures Search & Analysis

Molecule One Sp zoo

Overview

Related

Corporation Tree

Pipeline

Pipeline Snapshot as of 12 Mar 2026

Deal

Translational Medicine

Profit

Grant & Funding(NIH)

Investment

Financing