Regularisasi Regresi Cox dengan Adaptive-LASSO untuk Memprediksi Prognosis Pasien Adenokarsinoma Paru

Nafilah, Chafshoh (2025) Regularisasi Regresi Cox dengan Adaptive-LASSO untuk Memprediksi Prognosis Pasien Adenokarsinoma Paru. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5003211011-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only until 1 April 2027. Download (2MB) | Request a copy

Abstract

Adenokarsinoma paru adalah subtipe non-small cell lung cancer (NSCLC) yang paling umum dideteksi sebagai penyebab utama akibat kanker paru secara global. Meskipun pengobatan terus berkembang, tetapi masih perlu model prognostik yang lebih baik untuk kelangsungan hidup pasien. Penggabungan data klinis dan ekspresi gen menawarkan peluang untuk meningkatkan kelangsungan hidup dan strategi pengobatan pasien yang disesuaikan. Pendekatan regresi Cox proportional hazard dengan regularisasi penalti adaptive-LASSO diterapkan untuk mengatasi kompleksitas data, memungkinkan seleksi variabel yang parsimoni, konsisten, dan relevan terhadap prognosis pasien. Data yang digunakan berasal dari Gene Expression Omnibus (GEO) dengan nomor akses "GSE68571", setelah melalui preprocessing mencakup 79 observasi dengan 3.013 variabel yang memuat tujuh variabel klinis dan 3.006 ekspresi gen. Hasil analisis menunjukkan bahwa stadium kanker merupakan faktor klinis utama yang memengaruhi kelangsungan hidup pasien, dengan pasien stadium III memiliki laju kematian lebih cepat 14,806 kali dibandingkan stadium I. Model Cox-Straso menujukkan model paling optimal dengan C-index yang paling tinggi yaitu 0,866. Model optimal mengidentifikasi dua variabel signifikan yang mengintegrasikan variabel klinis Stadium (2) dan ekspresi gen PRKACB (Protein kinase cAMP-activated catalytic subunit beta) dengan estimasi hazard ratio 14,806 dan 0,183. Peningkatan ekspresi gen PRKACB sebesar 1 unit log2 (2^1) dapat memperlambat laju kematian pasien 0,183 kali, menunjukkan peran protektif. Penelitian ini menunjukkan peran penting pendekatan regularisasi yaitu LASSO dan adaptive-LASSO dalam mengatasi data berdimensi tinggi yang dapat mengintergrasikan variabel klinis dan ekspresi gen dalam pemahaman faktor prognostik adenokarsinoma paru serta berpotensi membantu pengembangan strategi perawatan yang lebih efektif untuk meningkatkan kelangsungan hidup pasien.
===================================================================================================================================
Lung adenocarcinoma is the most common subtype of non-small cell lung cancer (NSCLC) detected as the leading cause of lung cancer outcomes globally. Although treatments continue to evolve, there is still a need for better prognostic models for patient survival. Combining clinical and gene expression data offers an opportunity to improve survival and tailor patient treatment strategies. A Cox proportional hazard regression approach with Adaptive-LASSO penalty regularization was applied to address the complexity of the data, enabling selection of variables that are parsimonious, consistent, and relevant to patient prognosis. The data used came from the Gene Expression Omnibus (GEO) with access number “GSE68571”, after preprocessing included 79 observations with 3,013 variables containing seven clinical variables and 3,006 gene expressions. The results of the analysis showed that cancer stage was the main clinical factor affecting patient survival, with stage III patients having a 14.806 times faster death rate than stage I patients. The Cox-Straso model showed the most optimal model with the highest C-index of 0.866. The optimal model identified two significant variables integrating the clinical variables Stage (2) and PRKACB (Protein kinase cAMP-activated catalytic subunit beta) gene expression with an estimated hazard ratio of 14.806 and 0.183. Increasing PRKACB gene expression by 1 log2 (2^1) unit can slow down the patient mortality rate by 0.183 times, suggesting a protective role. This study demonstrates the important role of regularization approaches, namely LASSO and adaptive-LASSO, in dealing with high-dimensional data that can integrate clinical variables and gene expression in understanding prognostic factors of lung adenocarcinoma and potentially help develop more effective treatment strategies to improve patient survival.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Adenokarsinoma Paru, Prediksi Prognosis, Ekpresi Gen, Regresi Cox, Adaptive-LASSO, Lung Adenocarcinoma, Prognosis Prediction, Gene Expression, Cox Regression
Subjects:	H Social Sciences > HA Statistics > HA29 Theory and method of social science statistics H Social Sciences > HA Statistics > HA31.3 Regression. Correlation. Logistic regression analysis. H Social Sciences > HA Statistics > HA31.38 Data envelopment analysis. Q Science > QH Biology > QH426 Genetics
Divisions:	Faculty of Science and Data Analytics (SCIENTICS) > Statistics > 49201-(S1) Undergraduate Thesis
Depositing User:	Chafshoh Nafilah
Date Deposited:	31 Jan 2025 16:37
Last Modified:	31 Jan 2025 16:37
URI:	http://repository.its.ac.id/id/eprint/117380

Actions (login required)

View Item