Unsupervised learning is a specific machine learning strategy in which the computer program independently finds patterns and correlations in given data. Labeled training data is not necessary. The model trains without supervision.

For his model, the prizewinner has developed new methods for meaningful extracts, which the program outputs from various raw data representations of molecules, e.g. in the form of texts, graphs and point clouds. The computer model is of interest for the development of subsequent models to predict molecular properties or to generate new molecules with desired properties and thus for the pharmaceutical drug development.

Original publications of the author on this topic:

Winter R, Montanari F, Noé F, Clevert DA. Learning continuous and data-driven molecular descriptors by translating equivalent chemical representations. Chem Sci. 2018 Nov 19;10(6):1692-1701. doi: 10.1039/c8sc04175j. PMID: 30842833; PMCID: PMC6368215.

Clevert DA, Le T, Winter R, Montanari F. Img2Mol - accurate SMILES recognition from molecular graphical depictions. Chem Sci. 2021 Sep 29;12(42):14174-14181. doi: 10.1039/d1sc01839f. PMID: 34760202; PMCID: PMC8565361.

Kim PT, Winter R, Clevert DA. Unsupervised representation learning for proteochemometric modeling. Int J Mol Sci. 2021 Nov 28;22(23):12882. doi: 10.3390/ijms222312882. PMID: 34884688; PMCID: PMC8657702.

Further information:
https://www.fu-berlin.de/presse/informationen/fup/2024/fup_24_331-tiburtius-preise-2024/index.html
https://refubium.fu-berlin.de/handle/fub188/42166
https://datasolut.com/wiki/unsupervised-learning/