Caterina Graziani (University of Siena)
Nov 25, 2020 – 11:00 – 12:30 AM
Ribosome profiling (Ribo-seq profiling) is a powerful tool for studying the translational control of gene expression. The translational profile of an RNA sequence is estimated by counting the occurrences of its subsequences in ribosome footprints. The translational speed of each part of the sequence can be digitized, identifying it as faster (-1) or slower (+1) than the median of the sequence. We worked on a dataset of 49 RNA sequences, carrying out a statistical analysis of the occurrences of nucleotides, couples of nucleotides and codons in order to conjecture the existence (or not) of signals in the sequence which modulate the speed. To validate our hypothesis, experiments were carried out in which neural networks were trained to predict the speed of translation of subsequences extracted from our dataset. An attention mechanism allowed to identify the signals which contributed the most to the network decisions.
Valleriani, Angelo, and Davide Chiarugi. “A workbench for the translational control of gene expression.” bioRxiv (2020).