Paper

2025

Comprehensive Atlas of Cancer-Type-Specific Molecular Features from Comparative Analysis of TCGA Data

Published:

Author: Yeu-Guang Tung
Abstract: Cancer-type-specific molecular alterations often reflect the unique biological context of their tissue of origin and are more likely to represent relevant drivers rather than passenger events. We analyzed data from The Cancer Genome Atlas (TCGA) spanning 39 cancer types and 6 molecular platforms to create a comprehensive atlas of cancer-type-specific molecular features through unified comparative analyses. Simple nucleotide variation analysis characterized cancer-type-specific gene mutations and revealed heterogeneity among mutated genes within shared pathways. Copy number variation analysis characterized cancer-type-specific amplifications and deletions and demonstrated synergistic interactions between gene deletions and mutations. DNA methylation analysis identified candidate hypermethylated genes alongside well-established targets. Transcriptome profiling analysis revealed cancer-type-specific pathway enrichment reflecting tissue-of-origin functions or novel associations. Multiomics clustering analysis identified multi-cancer clusters and revealed consistent patterns across molecular platforms. These findings provide insights into cancer-type-specific molecular features and offer comprehensive visualizations as a reference resource for clinical application and hypothesis generation.

2023

Predicting Electron Distributions in Drug Molecules Using Density Functional Theory in Real Space

Published:

Author: Yeu-Guang Tung
Abstract: Electron distributions in drug molecules are crucial in drug design and can be predicted using density functional theory (DFT). We propose a comprehensive scheme to perform DFT calculations in real space instead of using the conventional orbital basis set. We detail the implementation of spherical space and basis, Pulay’s density mixing, Hamiltonian matrix construction, Kohn-Sham equation solution, and Bader’s charge analysis in discretized coordinate basis. We demonstrate that the demanding computation can be carried out in a highly parallel manner with simple codes by exploiting efficient algorithms in the PyTorch and NumPy packages. We perform calculations on the drug acyclovir and predict the electron distribution, partial charges, and energy levels in the molecule. The study may facilitate research in computational molecular science and structure-based drug design.