The analysis of biological sequences is one of the main research areas of Bioinformatics.
Sequencing data are the input for almost all types of studies concerning genomic as well as transcriptomic sequences, and sequencing experiments should be conceived specifically for each type of application.
The challenges posed by fundamental biological questions are usually addressed by firstly aligning or assemblying the reads produced by new sequencing technologies. Assembly is the first step when a reference sequence is not available. Alignment of genomic reads towards a known genome is fundamental, e.g., to find the differences among organisms of related species, and to detect mutations proper of the so-called "diseases of the
genome". Alignment of transcriptomic reads against a reference genome, allows to detect the expressed genes as well as to annotate and quantify alternative transcripts.
In this thesis we overview the approaches proposed in literature for solving the above mentioned problems. In particular, we deeply analyze the sequence assembly problem, with particular emphasys on genome reconstruction, both from a more theoretical point of view and in light of the characteristics of sequencing data produced by state-of-the-art technologies. We also review the main steps in a pipeline for the analysis of the transcriptome, that is, alignment, assembly, and transcripts quantification, with particular
emphasys on the opportunities given by RNA-Seq technologies in enhancing precision.
The thesis is divided in two parts, the first one devoted to the study of local assembly methods for Next Generation Sequencing data, the second one concerning the development of tools for alignment of RNA-Seq reads and transcripts quantification. The permanent theme is the use of paired reads in all fields of applications discussed in this thesis. In particular, we emphasyze the benefits of assemblying inserts from paired reads in a wide range of applications, from de novo assembly, to the analysis of RNA.
The main contribution of this thesis lies in the introduction of innovative tools, based on well-studied heuristics fine tuned on the data. Software is always tested to specifically assess the correctness of prediction. The aim is to produce robust methods, that, having low false positives rate, produce a certified output characterized by high specificity.
