POINT: A Platform for unraveling interaction mechanisms between drugs and complex Diseases

Since its inception, network pharmacology has made significant progress in analyzing drug–disease interactions through biological networks, particularly in interpreting the multi-component, multi-target mechanisms of traditional Chinese medicine (TCM). However, most current studies face clear limitations: (i) They often rely solely on protein–protein interaction networks, overlooking network specificity across physiological and pathological states, as well as the roles of key regulators such as transcription factors, enhancers, and non-coding RNAs. (ii) Traditional approaches frequently depend on node degree for target prioritization, which may miss functionally critical but low-degree nodes, highlighting the need for algorithmic improvements. (iii) Interpretation of predictions heavily relies on prior knowledge. While knowledge graphs can provide biological context and validation, this potential remains underutilized.

Recently, a team led by Yi Zhao from the Institute of Computing Technology, Chinese Academy of Sciences, published a study in Science Bulletin introducing the POINT platform (http://point.gene.ac/). By integrating multi-omics networks, advanced network topology, and deep learning algorithms, along with a comprehensive biomedical knowledge graph, POINT offers a robust solution to current challenges in network pharmacology.

According to Zihao He, the first author of the study, “POINT integrates 1,976 biological networks, including 1,830 single-layer and 146 multi-layer networks. These comprise 25 general networks, 818 single-cell-specific, 669 cell-line-specific, 281 tissue-specific, and 183 disease-specific networks, allowing users to flexibly combine networks based on their research needs.”

Yang Wu, corresponding author from the Institute of Computing Technology, Chinese Academy of Sciences, explained, “In terms of algorithms, POINT incorporates both network-based methods, such as Random Walk with Restart, and deep learning methods, such as BACPI. The POINT platform establishes an end-to-end workflow covering network integration, target prediction, functional enrichment, and drug–disease relationship inference, enabling one-stop analysis from data processing to insight generation.”

Jincheng Guo, co-corresponding author from Beijing University of Chinese Medicine, added, “POINT’s biomedical knowledge graph integrates four authoritative sources—DRKG (drug screening), PharmKG (drug mechanisms), HERB (TCM knowledge), and iKraph (biomedical discovery)—ensuring comprehensive coverage. It includes 864,058 entities and 5,913,654 relationships across 10 categories, such as compounds, TCM herbs, genes, functional pathways, and diseases. This rich repository enhances the interpretability and reliability of network pharmacology results and supports multi-dimensional association analysis.”

Yi Zhao, co-corresponding author from the Institute of Computing Technology, Chinese Academy of Sciences, concluded, “The launch of POINT marks a key step toward multi-dimensional integrated analysis in network pharmacology. It systematically addresses three major limitations: (i) By integrating multi-omics networks, it moves beyond reliance on static protein–protein interaction networks, better reflecting real biological contexts. (ii) It combines Random Walk with Restart and deep learning algorithms to reduce bias from degree-based node ranking, improving target prediction accuracy. (iii) Its extensive knowledge graph, with massive entities and relationships, provides traceable biological explanations for predictions.
POINT not only offers a powerful tool for target discovery and drug–disease relationship inference in drug development but also, with its multi-dimensional network and TCM entity coverage, lays a technical foundation for modern research on TCM formula mechanisms.”

The research was co-supervised by corresponding authors Yang Wu and Yi Zhao from the Institute of Computing Technology, Chinese Academy of Sciences, and Jincheng Guo from Beijing University of Chinese Medicine. Zihao He, a postdoctoral fellow jointly trained by Ningbo Second Hospital and the Institute of Computing Technology, Chinese Academy of Sciences, served as the first author, with Liu Liu and Dongchen Han from Beijing University of Chinese Medicine as co-first authors.

---

---