DrugGen 2: A disease-aware language model for enhancing drug discovery

Source

• Source name: arXiv
• Source kind: paper
• Authors / organization: Ali Motahharynia, Mohammadreza Ghaffarzadeh-Esfahani, Mahsa Sheikholeslami, Navid Mazrouei, Matin Irajpour, Yousof Gheisari, Hajar Sirous
• Venue: arXiv: q-bio.QM
• Published: 2026-07-09
• Collected: 2026-07-10
• URL: http://arxiv.org/abs/2607.08404v1
• PDF: https://arxiv.org/pdf/2607.08404v1
• DOI: UNKNOWN

Plain-Language Brief

A new generative model called DrugGen-2 has been developed to enhance drug discovery. It takes into account both disease ontology and target protein sequences, improving upon existing models in generating unique molecules with better predicted binding affinities for five protein targets related to diabetic nephropathy.

What the Document Says

• DrugGen-2 is a novel generative model for drug design.
• It considers both disease ontology and target protein sequences.
• The model was developed by fine-tuning a pre-trained GPT-2 model on a dataset of approved drugs linked to their diseases and targets.
• The development process involved supervised fine-tuning followed by reinforcement learning via group relative policy optimization (GRPO).
• Reward functions were used, optimizing for chemical validity, novelty, diversity, and high predicted binding affinity.
• DrugGen-2 outperformed baseline models (DrugGPT and DrugGen) on five protein targets related to diabetic nephropathy.
• It demonstrated a superior capacity to generate unique molecules.
• The generated molecules showed greater structural similarity to approved drugs.
• Improved predicted binding affinities were achieved across all targets.
• Molecular docking analyses supported these findings, identifying candidate ligands with strong binding potential.
• Examples of such compounds include those with predicted affinities exceeding that of reference drugs like enalapril for angiotensin-converting enzyme.

Key Concepts Mentioned

• DrugGen-2
• disease ontology
• target protein sequences
• pre-trained GPT-2 model
• supervised fine-tuning
• reinforcement learning
• group relative policy optimization (GRPO)
• reward functions
• chemical validity
• novelty
• diversity
• predicted binding affinity
• molecular docking
• candidate ligands
• diabetic nephropathy
• angiotensin-converting enzyme

Claims Made by the Document

• DrugGen-2 outperforms baseline models.
• DrugGen-2 generates unique molecules.
• DrugGen-2 exhibits greater structural similarity to approved drugs.
• Improved predicted binding affinities are achieved with DrugGen-2.

Evidence Used

• When evaluated on five protein targets...
• It demonstrated a superior capacity...
• The generated molecules showed greater structural similarity...
• Improved predicted binding affinities were achieved across all targets...

Methods / Approach

• Fine-tuning of pre-trained GPT-2 model
• Supervised fine-tuning
• Reinforcement learning via GRPO

Limitations Stated by the Authors

UNKNOWN

Limitations Noted by the Agent

Agent interpretation — not from the source.

The text does not provide information on potential limitations of the study.

Open Questions

• Were there any potential biases in the dataset used for training DrugGen-2?
• How was the effectiveness of DrugGen-2 tested beyond the five protein targets mentioned?

Notes for Future Q&A

May include interpretation. Labeled where it does.

The paper focuses on the development and evaluation of DrugGen-2 for a specific set of protein targets related to diabetic nephropathy. It would be important to consider potential biases in the dataset used for training, as well as testing its effectiveness beyond these targets.

Raw Abstract / Excerpt

Current computational approaches for drug design typically focus on generating molecules conditioned on specific targets or general molecular properties, often neglecting the influence of disease context on target behavior and therapeutic outcomes. To address this gap, we introduce DrugGen-2, a novel generative model that designs small molecules conditioned on both disease ontology and target protein sequences. DrugGen-2 was developed by fine-tuning a pre-trained GPT-2 model on a curated dataset of approved drugs linked to their diseases and targets, using a two-step strategy of supervised fine-tuning followed by reinforcement learning via group relative policy optimization (GRPO). This process was guided by reward functions optimizing for chemical validity, novelty, diversity, and high predicted binding affinity. When evaluated on five protein targets relevant to diabetic nephropathy, DrugGen-2 significantly outperformed baseline models (DrugGPT and DrugGen). It demonstrated a superior capacity to generate unique molecules, exhibited greater structural similarity to approved drugs, and achieved improved predicted binding affinities across all targets. Molecular docking analyses further supported these findings, identifying candidate ligands with strong binding potential, including compounds with predicted affinities (-9.917, -9.485, and -9.367) exceeding those of reference drugs such as enalapril for angiotensin-converting enzyme (-8.283). By integrating disease-specific context into molecular generation, DrugGen-2 advances AI-assisted drug discovery, offering a powerful tool for de novo design and drug repurposing that accounts for the complex interplay between diseases and molecular targets.