Algorithms
Browse retrosynthesis algorithms and their model versions
| Name | Description | Versions |
|---|---|---|
| Version | Instance Name | Description | Prediction Runs | Date Added |
|---|---|---|---|---|
| Version | Instance Name | Description | Prediction Runs | Date Added |
|---|---|---|---|---|
A standardized and flexible Retro* harness for different single-step predictors developed by Microsoft
Current state-of-the-art achieved by any version of this algorithm.
95% CI: [4.4%, 13.1%]
Top-1 Accuracy achieved by Syntheseus LocalRetro (v0.7.0)
95% CI: [14.4%, 26.9%]
Top-10 Accuracy achieved by Syntheseus LocalRetro (v0.7.0)
95% CI: [6.2%, 11.2%]
Top-1 Accuracy achieved by Syntheseus LocalRetro (v0.7.0)
95% CI: [15.4%, 22.4%]
Top-10 Accuracy achieved by Syntheseus LocalRetro (v0.7.0)
| version | slug | description | prediction runs | date added |
|---|---|---|---|---|
v0.7.0 | synt-lr-v0-7-0 | — | 8 runs | 2 days ago |
if you use this model in your work, cite it as
@Article{syntheseus_2024,
author ="Maziarz, Krzysztof and Tripp, Austin and Liu, Guoqing and Stanley, Megan and Xie, Shufang and Gaiński, Piotr and Seidl, Philipp and Segler, Marwin H. S.",
title ="Re-evaluating retrosynthesis algorithms with Syntheseus",
journal ="Faraday Discuss.",
year ="2025",
volume ="256",
issue ="0",
pages ="568-586",
publisher ="The Royal Society of Chemistry",
doi ="10.1039/D4FD00093E",
url ="http://dx.doi.org/10.1039/D4FD00093E",
}