Key Findings
RetroTAC serves as a fast prioritization tool for virtual PROTAC libraries by predicting a route-derived synthesizability score directly from molecular structures. In evaluations using PROTACs created by recombining warheads, linkers, and E3-ligase ligands, candidates with high predicted scores achieved a mean route-derived synthesizability of 0.73, compared to 0.31 for low-score candidates. Furthermore, 98.75% of high-score candidates were fully resolved under tested search settings, whereas only 51.00% of low-score candidates were resolved.
Technical Details
The system introduces a human-interpretable, PROTAC-oriented score that rewards balanced joining of molecular fragments and short synthetic depth. RetroTAC utilizes an ensemble of complementary machine-learning models to act as a proxy for explicit retrosynthesis planning. To support the development of this model, the researchers have openly released the RetroTAC ensemble model and nearly 20,000 PROTAC synthesis routes generated by ShallowTree.
Background & Context
PROTACs are heterobifunctional molecules consisting of a target-binding warhead and an E3-ligase ligand connected by a linker. While their modular architecture allows for the modulation of proteins difficult to drug with conventional inhibitors, their large and flexible structures create synthetic bottlenecks. Traditional synthetic-accessibility scores often overestimate difficulty by penalizing molecular complexity and overlooking the modular nature of PROTAC assembly. While retrosynthesis planning tools provide explicit routes, they are computationally expensive and conditional on factors such as reaction templates, search budgets, and stochastic seeds.
Significance & Outlook
By reframing route assessment as a single-molecule property prediction, RetroTAC avoids the need to call a planner during inference. This allows researchers to integrate synthetic accessibility filters directly into generative pipelines or virtual screening campaigns. The tool allows for the rapid filtering of vast virtual libraries, reserving computationally intensive explicit planning only for the most promising candidates.
Source: https://arxiv.org/html/2610.13639v1
Get our weekly technology intelligence — free
Receive an infographic that lets you judge at a glance whether each field’s analysis report is worth reading.
Subscribe Free — Weekly Tech Intelligence
By subscribing, you’ll receive Troy-Technical’s weekly technology intelligence newsletter.
- Your email and selected fields are used only to deliver the newsletter.
- We never share your information with third parties.
- You can unsubscribe anytime via the link in each email.
See our Privacy Policy for details.
Takes about a minute · Unsubscribe anytime
