Luca Semeraro
Signature development for the detection of Pulmonary Hypertension.
Rel. Mauro Gasparini, Matthieu Villeneuve. Politecnico di Torino, Corso di laurea magistrale in Ingegneria Matematica, 2021
|
PDF (Tesi_di_laurea)
- Tesi
Licenza: Creative Commons Attribution Non-commercial No Derivatives. Download (2MB) | Preview |
Abstract: |
This thesis follows and assembles what I did during a six-months traineeship at Actelion Janssen. Once the company's therapeutic areas of interest are briefly described, the emphasis in this work is on one of them: Pulmonary Hypertension (PH). Then, a WHO classification and available diagnostic tests for this disease are examined and two clinical studies in PH labelled TRACE and CIPHER are outlined: the first is a multicenter, double-blind, placebo-controlled, phase 4 study to evaluate the effect of Uptravi treatment on the daily life physical activity of patients with pulmonary arterial hypertension, while the second one is a prospective, multicenter study designed to identify a biomarker signature for the early detection of PH and another signature to distinguish two sub-classes of this pathology from the others. The problem presented in the latter belongs to the challenging field of genetic analysis and is here addressed by developing a process for the identification of blood-based biomarker signatures for any disease through the implementation on R of different statistical methods such as Gradient Boosting, Support Vector Machines, GLM with elastic-net regularization and resampling techniques like Cross-Validation and Nested Cross-Validation. Thus, after summarizing what is available in the literature concerning biomarkers involved in the processes and mechanisms of the disease under investigation, the approach is applied to the dataset extracted from the proof-of-concept study used to design CIPHER, to search for a PH signature with microRNAs as biomarkers. Moreover, the resulting model is evaluated through some metrics such as sensitivity, specificity, and precision, and it is analyzed in detail together with its most influential biomarkers and the results separately by the different groups of the disease classification. Finally, the signature is compared with the standard non-invasive diagnostic method by means of the 95% level Wilson confidence region of sensitivity and specificity, and possible ways to improve its performance are proposed in the conclusion. |
---|---|
Relators: | Mauro Gasparini, Matthieu Villeneuve |
Academic year: | 2020/21 |
Publication type: | Electronic |
Number of Pages: | 126 |
Subjects: | |
Corso di laurea: | Corso di laurea magistrale in Ingegneria Matematica |
Classe di laurea: | New organization > Master science > LM-44 - MATHEMATICAL MODELLING FOR ENGINEERING |
Aziende collaboratrici: | Janssen |
URI: | http://webthesis.biblio.polito.it/id/eprint/25836 |
Modify record (reserved for operators) |