Asaduzzaman, A. | Search bioRxiv

bioRxiv2025

Using The Cancer Genome Atlas from cBioPortal to Develop Genomic Datasets for Machine Learning Assisted Cancer Treatment

Predicting the impact of genetic mutations is crucial for understanding diseases like cancer. Polymorphism Phenotyping (PolyPhen) and Sorting Intolerant From Tolerant (SIFT) are key tools for assessing how amino acid substitutions affect protein function and mutation pathogenicity. To our knowledge, no ready-to-use genomic dataset exists for prediction models to identify potentially harmful mutations, which could support research and clinical decisions. This study develops genomic and non-genomic datasets using The Cancer Genome Atlas (TCGA) from cBioPortal and applies machine learning models to predict PolyPhen and SIFT scores. We explore three classification models: Random Forest (RF), Extreme Gradient Boosting (XGBoost), and an ensemble RF-XGBoost model. Experimental results show that genomic data yields more accurate predictions than non-genomic data. The ensemble RF-XGBoost model performs best on genomic data, achieving average accuracies of 88.43% for PolyPhen and 95.13% for SIFT, highlighting the potential of artificial intelligence in genetic mutation analysis for disease treatment.

genomics↗

bioRxiv2024

Machine Learning Approaches for Skin Neoplasm Diagnosis

Approaches for skin neoplasm diagnosis include physical exam, skin biopsy, lab tests of biopsy samples, and image analyses. These approaches often involve error-prone and time-consuming processes. Recent studies show that machine learning has promises to effectively classify skin images into different classes such as melanoma and melanocytic nevi. In this work, we investigate machine learning approaches to enhance the performance of computer-aided diagnosis (CADx) systems to diagnose skin diseases. In the proposed CADx system, generative adversarial network (GAN) is used to identify (and remove) fake images. Exploratory data analysis (EDA) is applied to normalize the original dataset for preventing model overfitting. Synthetic minority over-sampling technique (SMOTE) is employed to rectify class imbalances in the original dataset. To accurately classify skin images, the following four machine learning models are utilized: linear discriminant analysis (LDA), support vector machine (SVM), convolutional neural network (CNN), and an ensemble CNN-SVM. Experimental results using the HAM10000 dataset demonstrate the ability of the machine learning models to improve CADx performance in treating skin neoplasm. Initially, the LDA, SVM, CNN, and ensemble CNN-SVM show 49%, 72%, 77%, and 79% accuracy, respectively. After applying GAN, EDA, and SMOTE, the LDA, SVM, CNN, and ensemble CNN-SVM show 76%, 83%, 87%, and 94% accuracy, respectively. We plan to explore other machine learning models and datasets in our next endeavor.

bioengineering↗