Skripsi

Prediksi Penghambatan Senyawa Dalam Zingiber Officinale Terhadap Sel Kanker Kolorektal HT-29 Dengan Pendekatan Machine Learning dan Kemoinformatik = Prediction of Compound Inhibition in Zingiber Officinale Against HT- 29 Colorectal Cancer Cells Using Machine Learning and Chemoinformatics Approaches.

Latar Belakang Kanker kolorektal merupakan masalah kesehatan global dengan terapi konvensional yang sering terkendala resistensi dan efek samping. Zingiber officinale (jahe) mengandung senyawa bioaktif yang berpotensi sebagai agen antikanker, namun skrining senyawa potensial secara konvensional memakan waktu dan biaya. Pendekatan in silico dengan machine learning dan kemoinformatik dapat mempercepat proses identifikasi senyawa kandidat. Metode Penelitian analitik observasional dilakukan secara in silico. Data senyawa aktif terhadap sel HT-29 diambil dari basis data ChEMBL dan diproses menggunakan perangkat lunak DataWarrior. Sebanyak 14.250 senyawa digunakan untuk membangun model prediksi IC₅₀. Dua algoritma machine learning, yaitu Support Vector Machine Regression (SVR) dan Partial Least Squares (PLS), dibandingkan kinerjanya berdasarkan nilai R² testing. Model terbaik kemudian digunakan untuk memprediksi nilai IC₅₀ 11 senyawa bioaktif jahe yang diambil dari basis data KNApSAcK dan PubChem. Hasil Model SVR menunjukkan kinerja prediktif yang lebih unggul (R² testing = 0,656; akurasi 82,9%) dibandingkan PLS (R² testing = 0,466). Prediksi terhadap 11 senyawa jahe mengidentifikasi delapan senyawa dengan aktivitas penghambatan sangat kuat (good activity). Berdasarkan tinjauan literatur, Tectochrysin terpilih sebagai senyawa paling prospektif karena memiliki bukti ilmiah yang kuat dalam menghambat sel kanker kolon melalui induksi apoptosis. Kesimpulan Pendekatan machine learning dan kemoinformatik berhasil mengidentifikasi Tectochrysin dalam jahe sebagai penghambat sel HT-29 yang paling potensial, sehingga direkomendasikan untuk validasi lebih lanjut.
Kata Kunci: Zingiber officinale, Kanker Kolorektal, HT-29, Machine Learning, Kemoinformatik, IC₅₀, Tectochrysin


Introduction Colorectal cancer is a global health problem where conventional therapies are often hampered by resistance and side effects. Zingiber officinale (ginger) contains bioactive compounds with potential as anticancer agents. However, conventional screening for potential compounds is time-consuming and costly. An in silico approach using machine learning and chemoinformatics can accelerate the identification of candidate compounds. Method This observational analytical study was conducted in silico. Data on compounds active against HT-29 cells were obtained from the ChEMBL database and processed using DataWarrior software. A total of 14,250 compounds were used to build IC₅₀ prediction models. Two machine learning algorithms, Support Vector Machine Regression (SVR) and Partial Least Squares (PLS), were compared based on their R² testing values. The best-performing model was then used to predict the IC₅₀ values of 11 ginger bioactive compounds obtained from the KNApSAcK and PubChem databases. Results The SVR model showed superior predictive performance (R² testing = 0.656; accuracy 82.9%) compared to PLS (R² testing = 0.466). Prediction of the 11 ginger compounds identified eight compounds with very strong inhibitory activity (good activity). Based on a literature review, Tectochrysin was selected as the most prospective compound due to strong scientific evidence of its ability to inhibit colon cancer cells through apoptosis induction. Conclusion The machine learning and chemoinformatics approach successfully identified Tectochrysin in ginger as the most potential HT-29 cell inhibitor, recommending it for further validation.
Keywords: Zingiber officinale, Colorectal Cancer, HT-29, Machine Learning, Chemoinformatics, IC₅₀, Tectochrysin.