PGFS++: Molecular Property Improvement under Synthesis and Diversity Constraints

Source

• Source name: arXiv
• Source kind: paper
• Authors / organization: Boqiao Zhang, Godbless James, Sai Krishna Gottipati, Andrew Fitzgibbon
• Venue: arXiv: cs.LG
• Published: 2026-08-19
• Collected: 2026-08-20
• URL: http://arxiv.org/abs/2608.19121v1
• PDF: https://arxiv.org/pdf/2608.19121v1
• DOI: UNKNOWN
• Registry group: academic_apis_and_preprints
• Evidence role: canonical_academic_index
• Language: en
• Country / region: UNKNOWN

Plain-Language Brief

Researchers present PGFS++, an advancement of a previous method called PGFS for improving molecular properties like drug-likeness or binding affinity. The new method addresses limitations in PGFS by using trainable embedding lookup tables to represent reaction templates and second reactants more effectively. This improves the desired property while preserving output diversity. Additionally, PGFS++ introduces a reward-hacking failure mode prevention mechanism.

What the Document Says

• PGFS is a reinforcement learning method for molecular improvement with limitations due to indirect reactant selection.
• PGFS+ is an improvement of PGFS using trainable embedding lookup tables, a more effective scoring function, and RL algorithm.
• PGFS+ improves the desired property but exposes a reward-hacking failure mode.
• PGFS++ is a new framework that improves target properties while preserving output diversity by treating an input molecule as a starting point for synthesis and applying learned reaction templates with compatible in-stock building blocks.

Key Concepts Mentioned

• molecular property improvement
• drug discovery
• reinforcement learning
• policy gradient
• synthesis-aware
• PGFS
• PGFS+
• PGFS++
• reactant embedding
• scoring function
• RL algorithm
• reward-hacking failure mode

Claims Made by the Document

• PGFS+ significantly improves the desired property.
• PGFS++ preserves high output diversity.

Evidence Used

• Improving molecular properties is a recurring task in early-stage drug discovery
• PGFS uses reactant embedding prediction making reactant selection indirect
• PGFS+ represents reaction templates and second reactants by trainable embedding lookup tables
• Combined with a more effective scoring function and RL algorithm, PGFS+ improves the desired property
• PGFS++ treats an input molecule as the start of a forward-synthesis trajectory
• PGFS++ produces a molecule with improved target properties, an explicit synthesis route, and structural similarity to the input.

Methods / Approach

UNKNOWN

Limitations Stated by the Authors

UNKNOWN

Limitations Noted by the Agent

Agent interpretation — not from the source.

UNKNOWN

Open Questions

UNKNOWN

Notes for Future Q&A

May include interpretation. Labeled where it does.

UNKNOWN

Raw Abstract / Excerpt

Improving molecular properties, such as drug-likeness or binding affinity, is a recurring task in early-stage drug discovery. However, molecules optimized in an unconstrained chemical space have limited practical value if they cannot be synthesized. Policy Gradient for Forward Synthesis (PGFS) is a synthesis-aware reinforcement learning method for molecular improvement, but its use of reactant embedding prediction makes reactant selection indirect, which, as we show, limits learning effectiveness. We first develop PGFS+, in which reaction templates and second reactants are represented by trainable embedding lookup tables. Combined with a more effective scoring function and RL algorithm, PGFS+ significantly improves the desired property. However, it exposes a reward-hacking failure mode: a powerful reactant search can map diverse input molecules to the same high-reward magnet molecule, improving the reward while collapsing the output diversity. We therefore introduce PGFS++, a synthesis-aware reinforcement learning framework for input-specific molecular improvement. Given an input molecule, PGFS++ treats it as the start of a forward-synthesis trajectory, applies learned reaction templates with compatible in-stock building blocks, and produces a molecule with improved target properties, an explicit synthesis route, and structural similarity to the input. Experiments on molecular improvement tasks show that PGFS++ improves target properties while preserving high output diversity.