Napovedovanje odziva na zdravljenje metastatskega melanoma iz podatkov o izražanju genov

Vohar, Lea

| | SLO | ENG | Cookies and privacy

Bigger font | Smaller font

First page > Show document

Show document

Title:	Napovedovanje odziva na zdravljenje metastatskega melanoma iz podatkov o izražanju genov
Authors:	ID Vohar, Lea (Author) ID Štiglic, Gregor (Mentor) More about this mentor... ID Štajdohar, Miha (Comentor) ID Ausec, Luka (Comentor)
Files:	MAG_Vohar_Lea_2022.pdf (1,15 MB) MD5: 65060A1718E4B7A14135F76DF47E40D6
Language:	Slovenian
Work type:	Master's thesis/paper
Typology:	2.09 - Master's Thesis
Organization:	FZV - Faculty of Health Sciences
Abstract:	Uvod: Zaviralci imunskih kontrolnih točk so v zadnjem desetletju naredili izjemen napredek pri zdravljenju metastatskega melanoma, vendar je odzivnost na zdravljenje relativno nizka. Prepoznavanje biologije odziva in odpornosti na zdravljenje sta prednostni nalogi za optimizacijo izbire zdravil in izboljšanje rezultatov bolnikov. V okviru naše študije smo ocenili genski podpis IPRES za napovedovanje odziva na imunoterapijo. Metode: Izvedli smo statistično analizo kliničnih podatkov bolnikov z metastatskim melanomom in temeljne korake razvoja napovednega modela na transkiptomskih podatkih. Napovedne modele smo zgradili z metodo multiple logistične regresije, naključnega gozda in nevronsko mrežo. Modele smo ocenili z 20-kratno ponovitvijo vgnezdenega 5-kratnega sorazmernega prečnega preverjanja. Rezultati: Z uporabo podpisa IPRES kot vhodne spremenljivke napovednih modelov se je za najboljšega izkazal naključni gozd z rezultatom pri vrednosti AUC 0,65 (95 % IZ: 0. 65– 0.66). Z integracijo statistično značilnih genomskih podatkov smo vrednost metrike AUC povišali na 0,72 (95 % IZ: 0,71–0,72). Razprava in zaključek: Geni IPRES so bili izbrani kot diferencialni geni. Izkazalo se je, da diferenčnost izražanja genov med neodvisnima bazama podatkov iste vrste raka ni ponovljiva in da diferencialni geni kot predstavniki signalnih poti nimajo nujno zadostne napovedne moči. Potrdili smo pomembnost združevanja – omik in uporabo modelov strojnega učenja za doseganje natančnejših napovedi.
Keywords:	melanom, imunoterapija, zaviralci kontrolnih točk, strojno učenje
Place of publishing:	Maribor
Publisher:	[L. Vohar]
Year of publishing:	2022
PID:	20.500.12556/DKUM-83308
UDC:	616-006.81:575.113(043.2)
COBISS.SI-ID:	134718979
Publication date in DKUM:	21.12.2022
Views:	1146
Downloads:	94
Metadata:
Categories:	FZV
:	Copy citation

Average score:	(0 votes)
Your score:	Voting is allowed only for logged in users.
Share:

Hover the mouse pointer over a document title to show the abstract or click on the title to get all document metadata.

Licences

License:	CC BY-NC-ND 4.0, Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International

Link:	http://creativecommons.org/licenses/by-nc-nd/4.0/
Description:	The most restrictive Creative Commons license. This only allows people to download and share the work for no commercial gain and for no other purposes.
Licensing start date:	20.10.2022

Secondary language

Language:	English
Title:	Predicting tumor response in metastatic melanoma based on gene expression signatures
Abstract:	Introduction: Immune checkpoint inhibitors have made significant progress in metastatic melanoma treatment over the past decade, but the response rate to this treatment is relatively low. Understanding the biology of response and resistance to treatment are high priority tasks for drug selection optimisation and improved patient outcomes. In this study, we evaluated the IPRES gene signature for predictiveness of immunotherapy response. Methods: We carried out a statistical analysis of metastatic melanoma patient clinical data and the fundamental steps of predictive model development on transcriptomic data. Prediction models were built using multiple logistic regression, random forest and neural networks. They were evaluated by nested 5-fold cross-validation repeated 20 times. Results: Using IPRES signature as input, the random forest model showed best performance with AUC 0.65 (95% CI: 0.65–0.66). We were able to improve the AUC metric to 0.72 (95 % CI: 0.71–0.72) by integrating statistically significant genome data. Discussion and Conclusion: IPRES genes were selected as differential ones. It turned out that differential expression of genes between independent databases of the same cancer type data is not replicable and that differential genes as representatives of signalling pathways do not necessarily have sufficient predictive power. We have excludes the IPRES signature from potential biomarkers. We have, however, confirmed the importance of – omic data integration and the use of machine learning models for achieving more accurate predictions.
Keywords:	: melanoma, immunotherapy, immune checkpoint inhibitors, machine learning

Comments

Leave comment

You must log in to leave a comment.

Comments (0)

0 - 0 / 0

There are no comments!

Back