Machine-Learning-Based Identification of Key Feature RNA-Signature Linked to Diagnosis of Hepatocellular Carcinoma

Area of Science:

• Oncology
• Bioinformatics
• Machine Learning

Background:

• Hepatocellular carcinoma (HCC) is a leading cause of cancer mortality globally.
• Early and accurate diagnosis of HCC is critical for patient prognosis and treatment efficacy.
• Developing novel diagnostic tools for HCC is a significant unmet need.

Purpose of the Study:

• To develop a machine learning model for HCC diagnosis.
• To integrate differentially expressed RNA signatures with laboratory parameters.
• To create a novel RNA signature-based diagnostic model for HCC.

Main Methods:

• Utilized five machine learning classifiers: KNN, RF, SVM, LGBM, and DNNs.
• Trained and tested models on 187 and 80 samples, respectively.
• Included 22 features: clinical data and specific RNA expression levels (e.g., miR-1298, miR-1262, RAB11A, STAT1).

Main Results:

• The LGBM classifier achieved the highest accuracy (98.75%) in HCC prediction.
• LGBM outperformed other models including Random Forest (96.25%) and DNN (91.25%).
• The model demonstrated high predictive performance for hepatocellular carcinoma.

Conclusions:

• The developed machine learning model, integrating RNA signatures and clinical data, shows potential as a novel diagnostic tool for HCC.
• Specific RNA signatures (RAB11A, STAT1, ATG12, miR-1262, etc.) are key components of the diagnostic model.
• This approach offers a promising avenue for early and accurate HCC detection.