Computational Molecular Biology 2025, Vol.15, No.4, 171-182 http://bioscipublisher.com/index.php/cmb 18 0 from each other. Modeling, automated construction, and heterogeneous data fusion - these seemingly dull parts actually support the entire framework. We also wanted to see if it could truly serve scientific problems, such as identifying key proteins and recognizing functional modules. Therefore, topological analysis was introduced, treating the atlas as an experimental field in systems biology. The extent to which the model can explain has always been a pain point. We have added an "evidence path" in the reasoning stage, making the prediction no longer just about the result but also allowing us to see the process. On the other hand, we have also developed a practical system by ourselves, incorporating tens of thousands of protein interaction relationships, allowing researchers to directly query and analyze them. Even more surprisingly, through the prediction of the atlas, we really discovered and verified new protein-protein interactions, which indicates that it is not just a theoretical tool but can bring about new biological discoveries. It can be said that this work has opened up a gap in concept and built a step in practice, leaving room for more complex applications in the future, such as clinical diagnosis and drug design. Although we have already built the prototype of the knowledge graph of molecular interactions, it is still far from being "complete". The next step is to expand the scale - not only to look at proteins, but also to incorporate aspects such as transcriptional regulation and metabolic pathways, making the map more like a true molecular panorama. Of course, this also means that the data is more complex and the relationships are more chaotic, and our methods need to be refined again. Algorithmically, we also aim to take another step towards explainability, such as introducing the causal reasoning approach, so that the model can not only predict but also clearly explain "why". Knowledge graphs are not static either. We plan to attempt modeling in the time dimension to see how the interaction network changes during the disease process. The system should also be more flexible, allowing for direct invocation in R or Python. It would be best if biologists could use it without learning new systems. We also plan to carry out more cooperation in the future and apply the atlas to real research, such as assisting in the analysis of rare disease mechanisms or the screening of drug targets. Ideally, researchers can not only use it but also feed their new discoveries back into the graph, allowing the system to "grow knowledge" on its own. Overall, our paths can be roughly divided into three: larger, smarter, and more practical. This is just the beginning. There is still a long way to go. Acknowledgments I would like to express my heartfelt thanks to all the teachers who have provided guidance for this study. Conflict of Interest Disclosure The author affirms that this research was conducted without any commercial or financial relationships that could be construed as a potential conflict of interest. References Alocci D., Mariethoz J., Horlacher O., Bolleman J., Campbell M., and Lisacek F., 2015, Property graph vs RDF triple store: a comparison on glycan substructure search, PLoS ONE, 10(12): e0144578. https://doi.org/10.1371/journal.pone.0144578 Bahaj A., and Ghogho M., 2024, A step towards quantifying, modelling and exploring uncertainty in biomedical knowledge graphs, Computers in Biology and Medicine, 184: 109355. https://doi.org/10.1016/j.compbiomed.2024.109355 Clancy R., Ilyas I., and Lin J., 2019, Scalable knowledge graph construction from text collections, In: Proceedings of the Second Workshop on Fact Extraction and VERification (FEVER, pp.39-46. https://doi.org/10.18653/v1/d19-6607 Collura V., and Boissy G., 2007, From protein-protein complexes to interactomics, Sub-cellular Biochemistry, 43: 135-183. https://doi.org/10.1007/978-1-4020-5943-8_8 Feng Z., Shen Z., Li H., and Li S., 2022, e-TSN: an interactive visual exploration platform for target-disease knowledge mapping from literature, Briefings in Bioinformatics, 23(6): bbac465. https://doi.org/10.1093/bib/bbac465 Glen A., Deutsch E., and Ramsey S., 2025, PloverDB: a high-performance platform for serving biomedical knowledge graphs as standards-compliant web APIs, Bioinformatics, 2025: btaf380. https://doi.org/10.1101/2025.03.09.642156
RkJQdWJsaXNoZXIy MjQ4ODYzNA==