OpenNovelty.org

SYNC: Measuring and Advancing Synthesizability in Structure-Based Drug Design

ICLR 2026 Conference SubmissionAnonymous Authors
OpenReview Score: 6.7Download Report PDF
Structure Based Drug Design; Synthesizable Drug Design; Controllable Generation
Abstract:

Designing 3D ligands that bind to a given protein pocket with high affinity is a fundamental task in Structure-Based Drug Design (SBDD). However, the lack of synthesizability of 3D ligands has been hindering progress toward experimental validation; moreover, computationally evaluating synthesizability is a non-trivial task. In this paper, we first benchmark eight classical synthesizability metrics across 11 SBDD methods. The comparison reveals significant inconsistencies between these metrics, making them impractical and inaccurate criteria for guiding SBDD methods toward synthesizable drug design. Therefore, we propose a simple yet effective SE(3)-invariant \textit{\underline{SYN}thesizability \underline{C}lassifier} (SYNC) to enable better synthesizability estimation in SBDD, which demonstrates superior generalizability and speed compared to existing metrics on five curated datasets. Finally, with SYNC as a plug-and-play module, we establish a synthesizability classifier-driven SBDD paradigm through guided diffusion and Direct Preference Optimization, where highly synthesizable molecules are directly generated without compromising binding affinity. Extensive experiments also demonstrate the effectiveness of SYNC and the advantage of our paradigm in synthesizable SBDD. Code is available at \url{https://anonymous.4open.science/r/SYNC-C94D/}.

Disclaimer
This report is AI-GENERATED using Large Language Models and WisPaper (A scholar search engine). It analyzes academic papers' tasks and contributions against retrieved prior work. While this system identifies POTENTIAL overlaps and novel directions, ITS COVERAGE IS NOT EXHAUSTIVE AND JUDGMENTS ARE APPROXIMATE. These results are intended to assist human reviewers and SHOULD NOT be relied upon as a definitive verdict on novelty.
NOTE that some papers exist in multiple, slightly different versions (e.g., with different titles or URLs). The system may retrieve several versions of the same underlying work. The current automated pipeline does not reliably align or distinguish these cases, so human reviewers will need to disambiguate them manually.
If you have any questions, please contact: mingzhang23@m.fudan.edu.cn

Overview

Taxonomy

Core-task Taxonomy Papers
50
3
Claimed Contributions
27
Contribution Candidate Papers Compared
0
Refutable Paper

Research Landscape Overview

Core task: synthesizability evaluation in structure-based drug design. The field has evolved to address a central challenge in computational drug discovery—ensuring that computationally generated molecules can actually be made in the laboratory. The taxonomy reveals several complementary branches: some focus on scoring and prediction methods that estimate how easily a molecule can be synthesized (often using machine learning or rule-based heuristics), while others embed synthesizability constraints directly into generative design workflows. A third major direction leverages retrosynthesis planning to guide molecular design, ensuring that proposed structures have plausible synthetic routes. Additional branches cover computational frameworks that integrate these considerations into end-to-end pipelines, application-specific design efforts (for example targeting particular proteins or disease areas), enzyme engineering for biocatalytic synthesis, benchmarking studies that evaluate synthesizability metrics, and specialized topics such as DNA-encoded libraries. Together, these branches reflect a shift from purely affinity-driven design toward practical, synthesis-aware strategies. Recent work highlights contrasting philosophies: some methods prioritize rapid scoring to filter large virtual libraries (Synthetic Accessibility Scoring[14], FSscore[25]), while others tightly couple generative models with retrosynthetic feasibility checks (Amortized Tree Generation[18], Generate What You Make[5]). The original paper, SYNC[0], sits within the generative design branch that optimizes synthesizability during structure-based generation. It shares common ground with approaches like Synthesis-Driven 3D Design[15] and Multi-Reward Optimization[26], which balance binding affinity and synthetic accessibility in a unified objective. Compared to purely scoring-based methods, SYNC[0] and its neighbors aim to steer the generative process itself rather than post-hoc filtering, reflecting an emerging consensus that early integration of synthesis constraints yields more practical candidate molecules. Open questions remain around the trade-offs between computational cost, the fidelity of synthesizability proxies, and the diversity of generated chemical space.

Claimed Contributions

Comprehensive benchmark of synthesizability metrics for SBDD

The authors establish a new benchmark that evaluates eight classical synthesizability metrics across five curated datasets containing easy-to-synthesize and hard-to-synthesize molecules. This benchmark reveals significant inconsistencies between existing metrics and provides a comprehensive evaluation of 11 SBDD methods.

10 retrieved papers
SYNC: SE(3)-invariant synthesizability classifier

The authors introduce SYNC, a 3D-aware and SE(3)-invariant classifier for predicting molecular synthesizability. SYNC demonstrates superior generalizability and speed compared to existing metrics and is designed to be fast, differentiable, 3D structure-aware, and SE(3)-invariant.

7 retrieved papers
Synthesizability classifier-driven SBDD paradigm

The authors propose a paradigm that integrates SYNC as a plug-and-play module into SBDD methods using two approaches: guided diffusion and Direct Preference Optimization (DPO). This paradigm enables generation of highly synthesizable molecules while preserving binding affinity without requiring additional binding constraints.

10 retrieved papers

Core Task Comparisons

Comparisons with papers in the same taxonomy category

Contribution Analysis

Detailed comparisons for each claimed contribution

Contribution

Comprehensive benchmark of synthesizability metrics for SBDD

The authors establish a new benchmark that evaluates eight classical synthesizability metrics across five curated datasets containing easy-to-synthesize and hard-to-synthesize molecules. This benchmark reveals significant inconsistencies between existing metrics and provides a comprehensive evaluation of 11 SBDD methods.

Contribution

SYNC: SE(3)-invariant synthesizability classifier

The authors introduce SYNC, a 3D-aware and SE(3)-invariant classifier for predicting molecular synthesizability. SYNC demonstrates superior generalizability and speed compared to existing metrics and is designed to be fast, differentiable, 3D structure-aware, and SE(3)-invariant.

Contribution

Synthesizability classifier-driven SBDD paradigm

The authors propose a paradigm that integrates SYNC as a plug-and-play module into SBDD methods using two approaches: guided diffusion and Direct Preference Optimization (DPO). This paradigm enables generation of highly synthesizable molecules while preserving binding affinity without requiring additional binding constraints.

Table of Contents