Predicting RNA-small molecule interactions with deep learning for RNA-targeted drug discovery
Apply to this project here
- Sponsored by: Prof Dr. Niki Kilbertus in cooperation with the RNA Biology lab (Helmholtz Munich)
- Project lead: Dr. Ricardo Acevedo Cabra
- Scientific lead: Giulia Cantini, Tobias Bernecker, Samuele Firmani
- TUM Co-Mentor: Dr. Ricardo Acevedo Cabra
- Term: Winter semester 2024
- Application deadline: Sunday 21.07.2024
Background & Motivation
AI is revolutionizing the field of drug discovery in the pharmaceutical industry by offering effective means to improve and optimize many existing processes, from target discovery to the identification of candidate compounds.
Although the field has so far been mostly geared toward the development of protein-based drugs, RNAs are attracting increasing attention in research and industry as potent therapeutic vectors and druggable biomolecules. Similarly to proteins, RNAs can be targeted by small
molecules to modulate their function with the goal of treating target-related diseases. This dramatically expands the pool of biological entities that can be drugged, offering novel ways to treat human diseases.
However, to date, only a few compounds have been approved as RNA drugs, and much effort is needed to push the discovery of novel and potent RNA-based medications. In-silico methods powered by artificial intelligence can support this process by providing an alternative and cost-effective method for the identification of RNA-binding small molecules.
Description & Goals
In this project, we will leverage the latest deep learning architectures for RNA and chemical compounds to predict RNA-small molecules interactions. Our aim is to develop a trustworthy model to be employed in virtual screenings for the identification of compounds to prioritize as promising drug candidates for further testing.
The objectives of this project include:
- Building abstract representations of RNA and small-molecule entities by leveraging RNA and chemical language models (LMs) [1,2,3] and/or geometric deep learning [4].
- Combining the two representations and training a deep learning model based on them to predict binding affinities of RNA to small molecules using public data available in the R-SIM [5]
database. - Interpreting the model to extract information and properties of RNA-binding molecules.
Requirements
- Good programming skills (Python)
- Prior knowledge of machine/deep learning with experience in PyTorch
Apply to this project here
References
[1] Chen, J., Hu, Z., Sun, S., Tan, Q., Wang, Y., Yu, Q., Zong, L., Hong, L., Xiao, J., King, I., & Li, Y. Interpretable RNA Foundation Model from Unannotated Data for Highly Accurate RNA Structure and Function Predictions. bioRxiv (2022). doi.org/10.48550/arXiv.2204.00300
[2] Peni'c, R.J., Vlasic, T., Huber, R., Wan, Y., & Šikić, M. RiNALMo: General-Purpose RNA Language Models Can Generalize Well on Structure Prediction Tasks. arXiv, 2024.
doi.org/10.48550/arXiv.2403.00043
[3] Moret, M., Pachon Angona, I., Cotos, L. et al. Leveraging molecular structure and bioactivity with chemical language models for de novo drug design. Nat Commun 14, 114 (2023).
doi.org/10.1038/s41467-022-35692-6
[4] Atz, K., Grisoni, F. & Schneider, G. Geometric deep learning on molecular representations. Nat Mach Intell 3, 1023–1032 (2021). doi.org/10.1038/s42256-021-00418-8
[5] Krishnan, S. R.; Roy, A.; Gromiha, M. M. R-SIM: A Database of Binding Affinities for RNA-small Molecule Interactions. J. Mol. Biol. 2022. 167914. doi: 10.1016/j.jmb.2022.167914