A method for parsing clinical outcomes and combining them with phosphoproteomic and genomic data for predicting drug efficacy: application in hepatocellular carcinoma

Postgraduate Thesis uoadl:1320816 266 Read counter

Unit:
Διιδρυματικό ΠΜΣ Τεχνολογίες Πληροφορικής στην Ιατρική και τη Βιολογία
Library of the School of Science
Deposit date:
2015-02-06
Year:
2015
Author:
Αηδονόπουλος Ορφέας
Supervisors info:
Σταύρος Περαντώνης (PhD - Διευθυντής Ερευνών: ΕΚΕΦΕ Δημόκριτος), Λεωνίδας Αλεξόπουλος (Επίκουρος Καθηγητής: ΕΜΠ), Γεώργιος Βερνίκος (PhD - Ερευνητής)
Original Title:
A method for parsing clinical outcomes and combining them with phosphoproteomic and genomic data for predicting drug efficacy: application in hepatocellular carcinoma
Languages:
English
Translated title:
Aνάπτυξη μεθόδου για πρόβλεψη κλινικής αποδοτικότητας φαρμάκων μέσω ενσωμάτωσης γενομικών, κλινικών και φωσφοπρωτεομικών δεδομένων: Eφαρμογή στον ηπατικό καρκίνο
Summary:
Systems biology has become an essential component of drug discovery, attempting
to combine experimental data with computational modeling to capture different
levels of cellular function (such as signaling, transcription, regulation) and
integrate them in predictive models. These models are then used to best
understand the drug mode of action (MOA), identify new targets and predict
clinical drug efficacy and toxicity. In this study, we tried to identify
signaling pathways related to drug efficacy in one of the most lethal
malignancies worldwide, hepatocellular carcinoma (HCC). Particularly, gene
expression data were collected for various HCC cell lines treated with
anticancer compounds of known clinical efficacy. Each compound had been
categorized with a ‘pass’ or ‘fail’ label according to their success or failure
in human clinical trials. For labeling each drug we constructed a graphical
user interface that parses clinical trials databases for clinical outcomes
containing the respective compound. Thus, having available a dataset consisting
of labeled drugs as observations and genes as features a supervised learning
method was applied (feature selection) to identify genes predictive of the
drugs’ clinical efficacy. Finally, using the extracted data as an input to a
pathway construction algorithm, we were able to infer signaling networks on the
proteomic level that best fit the measured gene expression signatures. We
identified reactions and pathways playing an important role as accurate
predictors for the efficacy of nine drugs in HCC.
Keywords:
Machine learning, Signaling pathways, Pathway construction, Gene expression, Data analysis
Index:
Yes
Number of index pages:
1
Contains images:
Yes
Number of references:
40
Number of pages:
101
File:
File access is restricted only to the intranet of UoA.

document.pdf
2 MB
File access is restricted only to the intranet of UoA.