Page 117 - AIH-1-3

Artificial Intelligence in Health ISM: A new multi-view space-learning model

molecule is an important challenge in drug discovery, Further disclosure
known as “scaffold hopping.” In this context, we are
48
currently analyzing the fingerprints of the docking of The paper has been uploaded to or deposited in a
10 of 1000 of molecules to dozens of proteins, with preprint server (https://www.preprints.org/manuscript
protein-associated fingerprints forming the different /202402.1001/v3).
views of each molecule. The goal is to use the ISM- References
transformed fingerprints to predict scaffold-hopping
chemotypes. Given the enormous size of the dataset – 1. Cichocki A, Zdunek R, Phan AH, Amari S. Nonnegative
each fingerprint contains more than 100 binary digits – matrix and tensor factorizations: Applications to exploratory
the ILSM strategy is being evaluated as a possible way to multi-way data analysis and blind source separation. IEEE
reduce computational problems, as smaller sets of views Signal Process Mag. 2009;25:142-145.
can be analyzed on smaller subsets of observations doi: 10.1002/9780470747278
before integrating them in their entirety. 2. Perry R, Mischler G, Guo R, et al. mvlearn: Multiview machine
learning in python. J Mach Learn ℝes. 2020;22(109):1-7.
Acknowledgments
doi: 10.48550/arXiv.2005.11890
Our sincere thanks to Prasad Chaskar, Translational
Medicine Senior Expert Data Science Lead at Galderma, for 3. Argelaguet R, Velten B, Arnol D, et al. Multi‐omics factor
stimulating discussions, especially on potential limitations analysis-a framework for unsupervised integration of multi‐
omics data sets. Mol Syst Biol. 2018;14(6):e8124.
arising from missing views when training latent models
with multiple views. We also thank Philippe Pinel from doi: 10.15252/msb.20178124
the Center for Computation Biology, Mines Paris/PSL, and 4. Argelaguet R, Arnol D, Bredikhin D, et al. MOFA+:
Iktos SAS, Paris France, for discussions on addressing ISM A statistical framework for comprehensive integration of
calculation challenges in Computational Biology. multi-modal single-cell data. Genome Biol. 2020;21(1):111.
Funding doi: 10.1186/s13059-020-02015-1
5. Wu J, Lin Z, Zha H. Essential tensor learning for multi-
None.
view spectral clustering. IEEE Trans Image Process.
Conflict of interest 2019;28(12):5910-5922.
doi: 10.1109/tip.2019.2916740
Franck Augé and Galina Boldina are employees of Sanofi
and may hold shares and/or stock options in the company. 6. Guo W, Che H, Leung M. Tensor-based adaptive consensus
All other authors declare no conflicts of interest. graph learning for multi-view clustering. IEEE Trans
Consum Electron. 2024.
Author contributions doi: 10.1109/tce.2024.3376397
Conceptualization: Paul Fogel, George Luta 7. Li J, Gao Q, Wang Q, Xia W, Gao X. Multi-View Clustering
Investigation: Franck Augé, Galina Boldina via Semi-Non-Negative Tensor Factorization. arXiv
Writing-original draft: Paul Fogel, Christophe Geissler, [Preprint]; 2023.
Galina Boldina doi: 10.48550/arXiv.2303.16748
Writing-review & editing: George Luta, Christophe Geissler, 8. Wang S, Cao J, Lei F, Jiang J, Dai Q, Ling BW. Multiple
Franck Augé
kernel-based anchor graph coupled low-rank tensor
Ethics approval and consent to participate learning for incomplete multi-view clustering. Appl Intell.
2022;53(4):3687-3712.
Not applicable.
doi: 10.1007/s10489-022-03735-6
Consent for publication 9. Zhao W, Gao Q, Li G, Deng C, Yang M. One-Step Multi-
View Clustering Based on Transition Probability. arXiv
Not applicable. [Preprint]; 2024.
Availability of data doi: 10.48550/arXiv.2403.01460

The data used in this article and the ISM Jupyter Python 10. Ali W, Yang M, Ali M, Ud-Din S. Fuzzy model-based sparse
notebook can be downloaded from the Advestis part of clustering with multivariate t-mixtures. Appl Artif Intell.
Mazars GitHub repository (https://github.com/Advestis/ 2023;37(1):2169299.
adilsm). doi: 10.1080/08839514.2023.2169299

Volume 1 Issue 3 (2024) 111 doi: 10.36922/aih.3427