Supervision: Mohan Vamsi Nallapareddy,Francesco Craighero

Project type: Semester project (master)

Available

Background Ribosome profiling [1] enabled in vivo monitoring of translation, allowing us to measure codon-specific ribosome dwell times. Recently, computational methods exploited this new ribosome density data to model translation elongation dynamics through Machine Learning [2]. However, most of these existing apporaches do not take into account perturbation in the cellular state [3], such as nutrient availability.

  1. Ingolia, N. Ribosome profiling: new views of translation, from single codons to genome scale. Nat Rev Genet 15, 205–213 (2014). https://doi.org/10.1038/nrg3645
  2. Tian, T., et al. (2021). Full-length ribosome density prediction by a multi-input and multi-output model. PLOS Computational Biology, 17(3), e1008842. https://doi.org/10.1371/journal.pcbi.1008842
  3. Shao, B., Yan, J., Zhang, J., & Buskirk, A. R. (2023). Riboformer: A Deep Learning Framework for Predicting Context-Dependent Translation Dynamics (p. 2023.04.24.538053). bioRxiv. https://doi.org/10.1101/2023.04.24.538053

The project. In this project we want to model the ribosome density profile through Variational AutoEncoders (VAE) [4]. Initally, we will evaluate the effectiveness of VAE in reconstructing the ribosome profile in one condition. Then, we will extend the decoder to take into account also nutrient availability (see [5] for an example where the condition is employed for voice conversion).

  1. I suggest to have a look at this book, chapter 21.
  2. Ding, S., Gutierrez-Osuna, R. (2019) Group Latent Embedding for Vector Quantized Variational Autoencoder in Non-Parallel Voice Conversion. Proc. Interspeech 2019, 724-728, doi: 10.21437/Interspeech.2019-1198

Expected Outcome. The candidate is expected to deliver a working implementation of a VAE modeling ribosome profiling. Depending on the time constraints and on the motivation of the candidate, we have many possibile extensions to the main outcome of the project.

Your profile. The main requirements include a strong understanding of Deep Learning and proficiency in Pytorch development. Prior familiarity with VAE and Pytorch Lightning is a plus. A background in biology is not necessary.

Application. If you are interested, you can get in touch with Francesco Craighero. Please provide your CV or a brief description of your profile.