Veronica Lachi (University of Siena)
Nov 25, 2020 – 11:00 – 12:30 AM
Ribosome profiling — which nowadays is the best tool to investigate the gene expression control — allows to determine the speed of translation of an RNA sequence, by analysing ribosome footprints. Among samples from the same RNA sequence, a higher occurrence in ribosomes suggests a slower translation. Discretizing the speed profile allows to assign a +1 (slow) or -1 (fast) label to each nucleotide.
In this thesis, a statistical method was devised to analyse 49 labelled RNA sequences from the E. Coli genome. In addition, four neural network models — which have been trained to predict the translational speed of codons — are evaluated in order to determine which one is the most suitable for this task.
The input data consists of the codon for which we want to predict the translational speed and a context. Further analysis allows to find the context’s size that leads to the most accurate predictions.