Anna Visibelli (DIISM, University of Siena)

Jun 5, 2019 – 11:00 AM

DIISM, Artificial Intelligence laboratory (room 201), Siena SI

Description

Nowadays, determining the 3D structure of proteins is one of the most challenging tasks in computational biology, essential for understanding protein functions. Thanks to experimental tests, protein conformations can be mainly determined by the sequence of its amino acids [C.B. Anfinsen, Principles governing the folding of protein chains, Science], a fundamental hypothesis in the development of protein folding prediction techniques for several decades. Although gradual developments have been made in predicting proteins 3D structure, the results obtained are generally of lower quality than the ones obtained for protein secondary structure, which provides a significant first step toward tertiary structure prediction, also yielding information about protein activity and function. Moreover, based on the intuition that signals should exist at the secondary structure boundaries, we performed a statistical analysis of their relative presence in the vicinity of alpha-helices. Moreover, recent developments in protein folding techniques have seen the use of neural networks for protein structures prediction. For this reason, we implemented a LSTM neural network to highlight the importance of the border amino-acids in determining the alpha-helix formation.