Nona Edizione della Giornata Toscana di Bioinformatica e Systems Biology,
15-16 Settembre 2022
Auala Gerace, Dipartimento di Informatica dell'Università di Pisa, largo Bruno Pontecorvo 3, Pisa
|8:30-8.45||Opening and welcome remarks|
Explainable AI and bioinformaticsDr Filippo Geraci, IIT-CNR|
Every day, without asking ourselves many questions, we rely on artificial intelligence for new tasks on which, increasingly, our lives depend. Among these many tasks, algorithms support clinicians in making diagnoses and implementing the so-called personalized medicine. Understanding the reasons behind algorithms' choices not only would encourage patient confidence, but may help designers to validate the correctness of models. Explainable artificial intelligence, however, requires new algorithms and analysis techniques that are the focus of research in the field. In this tutorial we will discuss the implications and challenges of explainable artificial intelligence in bioinformatics by giving some practical examples.
|10:00-10:15||Presentazione del Master in Bioinformatica e Data ScienceProf Moreno Falaschi, Univ. di Siena|
Reaction Systems processes for modeling biological systemsProf Linda Brodo, Univ. di Sassari|
Reaction Systems (RSs) provide a computational (mechanistic) framework inspired by ‘natural computing’. A RS pairs a set of entities with a set of reactions over them. Entities can be used to enable or inhibit each reaction, and are produced by reactions. A RS can interact with the environment by means of an external context sequence to simulate in-silico biological experiments. RSs can model several computer science frameworks as well as biological systems. However the basic computational mechanism of RSs abstracts from several features of biochemical systems which reduces somewhat their expressivity.
In this tutorial we will define a computational framework which represents RSs as process algebras. This allows us to introduce several (quantitative) extensions of RSs. We will also show how it is possible to verify properties of biological systems modeled by RSs processes. We will briefly describe a prototype implementation of our framework in Prolog. The interpreter represents a rapid prototyping tool which simplifies modifications to accommodate new extensions.
Genomic Sequences Alignments with Dynamic ProgrammingProf. Nadia Pisanti, Univ. di Pisa|
In bioinformatics, sequence alignment is a way of arranging two pr more sequences of DNA, RNA, or proteins to highlight regions of high similarity and regions where variants occur, both being consequence of structural or evolutionary relationships between the sequences.
Computational approaches to pairwise sequence alignments can seek for global alignments (both sequences have to be involved entirely), local alignments (fragments of the input with high similarity are sought), or semi-global alignment (one of the two input sequences has to be entirely aligned to a fragment of the other one). For all these problem variants a powerful and elegant dynamic programming method can be applied. In this tutorial we introduce this method framing the techniques in its algorithmic theory.
Molecular generative Graph Neural Networks for drug discoveryProf. Monica Bianchini, Univ. di Siena|
Drug discovery is a fundamental and ever-evolving field of research. The design of new candidate molecules requires large amounts of time and money, and computational methods are increasingly being used to reduce these costs. Machine learning methods are ideal for massive design of potential new candidate molecules, which are naturally represented as graphs and, indeed, the generation of molecular graphs is one of the most promising applications of recent deep learning techniques.
In this tutorial, after some introductory notions on Graph Neural Networks (GNNs), a sequential molecular graph generator, based on a set of GNN modules, is presented. At each step, a node or a group of nodes is added to the graph, along with its connections. The modular architecture simplifies the training procedure, also allowing an independent retraining of a single module. Sequentiality and modularity make the generation process interpretable. The use of GNNs maximizes the information in input at each generative step, which consists of the subgraph produced during the previous steps. Experiments of unconditional generation on the QM9 and Zinc benchmark datasets show that the model can generalize molecular patterns seen during the training phase, without overfitting.
Genomica dei tumoriDr. Romina D'Aurizio, IIT-CNR|
Technological innovation and rapid reduction in sequencing costs have enabled the genomic profiling of hundreds of cancer-associated genes as a component of routine cancer care. A key mutational process in cancer involves somatic Copy Number Alterations (CNA) that are associated with aneuploidy, which is a context-dependent, cancer-type-specific oncogenic event influencing tumor proliferation and immune response escaping. In this tutorial, we will focus on computational methods for dealing with cancer complexity starting from real sequencing datasets to extract meaningful molecular features that have a direct clinical relevance in Precision Oncology both as prognostic markers and as potential therapeutic targets.
Fundamentals of microbiome data analysisProf. Roberto Marangoni, Univ. di Pisa|
Despite the various experimental approaches, microbiome data analysis is mainly based on concepts and statistical tools derived from classical ecology. The starting point is always represented by an abundance table where the quantitative presence of each species in different samples is recorded. The abundance table is usually processed to make the different samples comparable to each other, with the final goal to discover possible metabolic/environmental factors able to alter the microbiome composition. Several ordination methods have been developed to deeply investigate possible inter-specific interactions. This tutorial will illustrate the problems underlying a typical microbiome data analysis pipeline from a biological point of view, discussing in particular the issue of non-linear effects involving low-abundance species.
|9:30-9:45||Opening of BH9 Local Organizing CommitteeGeraci F., Marangoni R., Pellegrini M., Pisanti N.|
|9:45-10:00||DaScH Lab: accelerating biologicals discovery through data scienceG. Maccari (TLS)|
|10:00-10:15||The Computational and Translational Genomics Lab's activities (IIT-CNR)R. D'Aurizio (CNR)|
|10:15-10:30||Biosystems modelling group @Unipi: from gene regulation networks to protein structure and interactionsP. Milazzo (Unipi)|
|10:30-10:45||Malignant pleural mesothelioma: clinicopathologic and differential gene expression analysis for mesothelial hyperplasia differential diagnosisD. Rosati (Unisi)|
|10:45-11:00||Less is enough: identification of actionable cancer neoepitopes using just tumor RNA-seqD. Tatoni (CNR)|
|11:30-11:45||miXer: a Machine-learning method to detect genomic Imbalances exploiting X chromosome Exome Reads E. Ceroni (Unisi, CNR)|
|11:45-12:00||The haplotype-phased genome of Ficus carica: A new genomic tool to elucidate the heterozygosity conditionG. Usai (Unipi)|
|12:00-12:15||Novel gene formation: the phenomenon of exapted transposable elements in a large genomeM. Ventimiglia (Unipi)|
|12:15-12:30||MeStudio, An epigenomic analysis tool for crossing methylation and genomic features in bacteriaI. Passeri (Unifi)|
|12:30-12:45||A multi-omics approach to liquid biopsies in Soft Tissue SarcomasA. Bernini (Unisi)|
|14:30-14:45||Bioinformatic survey of X/Gly genomic variantsP. Bongini (Unisi)|
|14:45-15:00||Bioinformatics analyses provide new insights into protein/small-molecule bindingA. Trezza (Unisi)|
|15:00-15:15||Modeling the metabolic consequences of antimicrobial exposure in Escherichia coliT. Vasquez (Unifi)|
|15:15-15:30||Computational analysis of multi-parametric flow cytometric data to dissect antigen-specific B cell subsets elicited by the BNT162b2 mRNA COVID-19 vaccine S. Lucchesi (Unisi)|
|16:00-16:15||Perturbation of microRNA regulatory networks design different Waddington's epigenetic landscapes for neuroendocrine tumorsE. Luzi (Virtual lab)|
|16:15-16:30||Signal processing and machine learning tools for the analysis of biological data in personalized medicineL Billeci (CNR)|
Poster sessionE. Riccucci (Sant'Anna), B. Gianibbi (Unisi)|
Poster 1: Pan-genome of a MAGIC maize population's founder lines: a tool for photosynthesis-related traits identification
Poster 2: Ab initio folding methods for accurate 3D modeling of HGD mutants in Alkaptonuria disease