Perbandingan Random Forest Dan Xgboost Untuk Prediksi Gen Essensial Drosophila Melanogaster

Saputri, Laili Rahmawati Ine (2026) Perbandingan Random Forest Dan Xgboost Untuk Prediksi Gen Essensial Drosophila Melanogaster. Other thesis, Institit Teknologi Sepuluh Nopember.

Text 5002221146-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (4MB) | Request a copy

Abstract

Gen esensial berperan penting dalam kelangsungan hidup organisme. Hilangnya fungsi gen tersebut dapat menyebabkan kematian atau penurunan kebugaran. Identifikasi gen esensial melalui metode eksperimental seperti knockout screening masih menghadapi kendala biaya dan waktu. Penelitian ini menggunakan pendekatan machine learning untuk memprediksi esensialitas gen pada Drosophila melanogaster menggunakan Random Forest dan XGBoost. Data esensialitas diperoleh dari OGEE dan DEG, sedangkan sekuens DNA dan protein diperoleh dari FlyBase. Sebanyak 7.524 gen direpresentasikan menggunakan 1.434 fitur intrinsik yang mencakup komposisi nukleotida, asam amino, dipeptida, k-mer, serta berbagai sifat fisikokimia protein. Seleksi fitur dilakukan menggunakan ElasticNetCV, Mutual Information, dan Variance Inflation Factor, sedangkan ketidakseimbangan kelas ditangani menggunakan SMOTE. Evaluasi dilakukanmelalui nested cross-validation dengan 10 outer fold dan 5 inner fold, optimasi hiperparameter, serta penentuan ambang klasifikasi. Random Forest memperoleh precision 0,1866, recall 0,2812, F1-score 0,2189, ROC-AUC 0,7374, dan PR-AUC 0,1560. XGBoost memperoleh precision 0,2004, recall 0,2176, F1-score 0,1905, ROC-AUC 0,7283, dan PR-AUC 0,1731. Random Forest memiliki recall, F1-score, dan ROC-AUC lebih tinggi, sedangkan XGBoost memiliki precision dan PR-AUC lebih tinggi. Analisis SHAP menunjukkan bahwa ENC proxy dan beberapa fitur komposisi dipeptida memberikan kontribusi terhadap prediksi kedua model.
===========================================================================================================================
Essential genes play a crucial role in the survival of organisms, as the loss of function of these genes can lead to death or reduced fitness. Identifying essential genes through experimental methods such as knockout screening still faces cost and time constraints. This study employs a machine learning approach to predict gene essentiality in *Drosophila melanogaster* using Random Forest and XGBoost. Essentiality data were obtained from OGEE and DEG, while DNA and protein sequences were obtained from FlyBase. A total of 7,524 genes were represented using 1,434 intrinsic features, including nucleotide composition, amino acids, dipeptides, k-mers, and various physicochemical properties of proteins. Feature selection was performed using ElasticNetCV, Mutual Information, and Variance Inflation Factor, while class imbalance was addressed using SMOTE. Evaluation was conducted via nested cross-validation with 10 outer folds and 5 inner folds, hyperparameter optimization, and classification threshold determination. Random Forest achieved a precision of 0.1866, a recall of 0.2812, an F1-score of 0.2189, an ROC-AUC of 0.7374, and a PR-AUC of 0.1560. XGBoost achieved a precision of 0.2004, a recall of 0.2176, an F1-score of 0.1905, an ROC-AUC of 0.7283, and a PR-AUC of 0.1731. Random Forest demonstrated higher recall, F1-score, and ROC-AUC, whereas XGBoost achieved higher precision and PR-AUC. SHAP analysis showed that the ENC proxy and several dipeptide composition features contributed substantially to the predictions of both models.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Drosophila melanogaster, Gen esensial, Random Forest, SHAP, SMOTE, XGBoost
Subjects:	Q Science > Q Science (General) > Q325.5 Machine learning. Support vector machines.
Divisions:	Faculty of Science and Data Analytics (SCIENTICS) > Mathematics > 44201-(S1) Undergraduate Thesis
Depositing User:	Laili Rahmawati Ine Saputri
Date Deposited:	04 Aug 2026 01:33
Last Modified:	04 Aug 2026 01:33
URI:	http://repository.its.ac.id/id/eprint/142510

Actions (login required)

View Item