Paper List

Journal: ArXiv Preprint
Published: Unknown
BioinformaticsDeep Learning

Contrastive Deep Learning for Variant Detection in Wastewater Genomic Sequencing

Georgia State University, Atlanta, Georgia, USA

Adele Chinda, Richmond Azumah, Hemanth Demakethepalli Venkateswara
Figure
Figure
Figure
Figure

The 30-Second View

IN SHORT: This paper addresses the core challenge of detecting viral variants in wastewater sequencing data without reference genomes or labeled annotations, overcoming issues of high noise, low coverage, and fragmented reads.

Innovation (TL;DR)

  • Methodology First comprehensive application of VQ-VAE with EMA quantization to wastewater genomic surveillance, achieving 99.52% token-level reconstruction accuracy with 19.73% codebook utilization.
  • Methodology Integration of masked reconstruction pretraining (BERT-style) maintaining ~95% accuracy under 20% token corruption, enabling robust inference with missing/low-quality data.
  • Methodology Contrastive fine-tuning with varying embedding dimensions showing +35% (64-dim) and +42% (128-dim) Silhouette score improvements, establishing representation capacity impact on variant discrimination.

Key conclusions

  • VQ-VAE achieves 99.52% mean token-level accuracy and 56.33% exact sequence match rate on SARS-CoV-2 wastewater data with 100,000 reads.
  • Contrastive fine-tuning improves clustering performance by +35% (0.31→0.42) with 64-dim embeddings and +42% (0.31→0.44) with 128-dim embeddings.
  • The framework maintains efficient codebook utilization (19.73%, 101 of 512 codes active) while providing robust performance under data corruption.
Background and Gap： Traditional reference-based variant calling pipelines struggle with novel mutations, low-frequency variants in mixed populations, and require substantial computational resources, while continuous VAEs suffer from posterior collapse and are unsuitable for discrete mutational events.

Abstract: Wastewater-based genomic surveillance has emerged as a powerful tool for population-level viral monitoring, offering comprehensive insights into circulating viral variants across entire communities. However, this approach faces significant computational challenges stemming from high sequencing noise, low viral coverage, fragmented reads, and the complete absence of labeled variant annotations. Traditional reference-based variant calling pipelines struggle with novel mutations and require extensive computational resources. We present a comprehensive framework for unsupervised viral variant detection using Vector-Quantized Variational Autoencoders (VQ-VAE) that learns discrete codebooks of genomic patterns from k-mer tokenized sequences without requiring reference genomes or variant labels. Our approach extends the base VQ-VAE architecture with masked reconstruction pretraining for robustness to missing data and contrastive learning for highly discriminative embeddings. Evaluated on SARS-CoV-2 wastewater sequencing data comprising approximately 100,000 reads, our VQ-VAE achieves 99.52% mean token-level accuracy and 56.33% exact sequence match rate while maintaining 19.73% codebook utilization (101 of 512 codes active), demonstrating efficient discrete representation learning. Contrastive fine-tuning with different projection dimensions yields substantial clustering improvements: 64-dimensional embeddings achieve +35% Silhouette score improvement (0.31→0.42), while 128-dimensional embeddings achieve +42% improvement (0.31→0.44), clearly demonstrating the impact of embedding dimensionality on variant discrimination capability. Our reference-free framework provides a scalable, interpretable approach to genomic surveillance with direct applications to public health monitoring.