Diffusion Deep Learning for Genome Variant Data
Abstract
Diffusion models have recently achieved remarkable success in generative modeling, particularly strong performance on continuous data in fields such as image and video generation. However, their application to discrete data remains under-explored. In this thesis, we explored the application of one-dimensional UNet-based diffusion models for generating diploid genomic sequences encoded as [0,1,2], using data from 2067 individuals. We evaluated model behavior across synthetic, mixed, and real genomic datasets to assess how increasing data complexity affects learning dynamics. The results showed that the model successfully learned simple synthetic structures, achieving near-perfect reconstruction and preserving positional patterns. However, its performance deteriorated substantially on realistic genomic data, where generated samples exhibited smoothing between genotype levels, weaker positional fidelity, and reduced preservation of discrete structure. We further explored mixed dataset and applied a range of architectural and optimization modifications, including conditioning strategies, augmentation methods, checkpoint fine-tuning, and learning-rate scheduling. While some training strategies improved stability and reduced Wasserstein distance, none fully resolved the difficulty of modeling complex discrete genomic distributions. Our findings indicate that standard continuous diffusion architectures face significant challenges when applied directly to discrete,…
Read the paper
Related studies
- Learning from First-Generation Qualitative Approaches in the IS Discipline: An Evolutionary View and Some Implications for Authors and Evaluators (PART 1/2)
- Commentary on the clinical and preclinical dosage limits of interstitially administered magnetic fluids for therapeutic hyperthermia based on current practice and efficacy models
- Prosthesis: New Technological Opportunities and Innovative Biomedical Devices
- The history and future of neural modeling for cochlear implants
- Ryanodine receptors: waking up from refractoriness
- Novel Transgenic Rabbit Model Sheds Light on the Puzzling Role of Matrix Metalloproteinase-12 in Atherosclerosis
Go deeper in Evidence Hub
Evidence Hub adds a study data sheet (design, sample size, evidence level, key numbers), an AI summary and a chat grounded in the paper's text.