Blog

Research activities2025.07.01

Publication Announcement: Development of a Comprehensive Molecular Docking Approach for Toxicity Target Identification Using the AlphaFold-Predicted Structural Proteome!

Our Latest Research Published in The Journal of Toxicological Sciences!

We are pleased to announce that our recent research has been published in the peer-reviewed international journal The Journal of Toxicological Sciences (Vol. 50, No. 7), issued by the Japanese Society of Toxicology.

Paper Title:
“Comprehensive molecular docking on the AlphaFold-predicted protein structure proteome: identifying target protein candidates for puberulic acid”
(Authors: Teppei Hayama, Rin Sugawara, Ryo Kamata, Masakazu Sekijima, Kazuki Takeda)

▶ View the full article here


Background

Identifying the molecular targets of toxic compounds remains one of the major challenges in toxicology.
While determining whether a chemical is toxic can often be achieved through animal exposure studies or in vitro assays, understanding how toxicity occurs—specifically, which proteins the compound interacts with—is far more complex.

For instance, puberulic acid, a harmful compound recently linked to severe kidney injury in the red yeast rice supplement incident, has been shown to cause nephrotoxicity in rodent studies. However, its exact mode of action and molecular targets remain largely unknown.


Our Approach

To address this, we developed a novel computational pipeline that leverages the AlphaFold2-predicted human proteome to perform comprehensive molecular docking across more than 20,000 proteins.
This large-scale in silico screening enables the prediction of potential protein targets without relying on pre-existing toxicity or interaction databases.
We applied this approach to puberulic acid to identify candidate target proteins involved in its toxicity.


Key Highlights

  • High-throughput molecular docking: Parallel GPU-based docking calculations performed on ~20,000 human and mouse proteins.

  • Integration of structural science and toxicology: Utilizing AlphaFold2 models to predict novel interactions beyond existing datasets.

  • Candidate target identification: Suggested the sodium/myo-inositol cotransporter (SMIT2), involved in renal osmotic regulation, as a promising target.

  • Introduction of “Binding Proteomics”: Proposed a new framework for proteome-wide in silico screening of chemical-protein interactions.


Why This Matters

Our Binding Proteomics approach opens new possibilities for:

  • Predicting the toxicity of novel compounds or those with unknown mechanisms,

  • Assessing species-specific sensitivity, including wildlife and non-model species.

Importantly, this workflow is fully open-source and available on GitHub, making it reproducible and accessible for the research community.


📢 This study provides a new computational strategy for chemical risk assessment in pharmaceuticals and environmental toxicology.
Moving forward, our lab aims to expand this approach to multi-species proteomes and integrate it with AI-based prediction models.

Latest Blog