Interpretable AI for genomic sequence analysis

SWAEV Genomics is building a cloud based API that lets researchers and biotech companies to analyse DNA sequences with models that explain why they make predictions. Unlike traditional "black box models", our architecture reveals the biological features (e.g. helical periodicity, regulatory motifs) driving each result.

SWAEV can also operate as a specialized expert within larger Mixture of Experts (MoE) systems, acting as an "interpretability layer" that grounds foundational LLMs by explaining the reasoning behind complex DNA analysis results.

Key use cases:

• Regulatory element discovery
• Variant effect prediction
• Synthetic promoter design
• "Why" layer interpretability for MoE architectures

We are currently seeking early access partners to validate our platform on proprietary datasets.

Our platform extends the open source SPECTRE‑Wave architecture with novel extensions designed specifically for DNA (vocabulary size = 4). Unlike standard Transformer models which use O(n²) attention algorithms that struggle with the massive scale of genomic data, our model replaces attention with wave propagation. This achieves a highly scalable O(n log n) complexity, enabling multi-million base pair context windows while maintaining full interpretability.

We have validated the architecture on E. coli, yeast, and human chromosome 22. During testing, the model autonomously discovered DNA's helical periodicity (10.5bp)—an emergent result that was not programmed in.

To scale beyond our current prototype hardware to full human genomes, we are actively seeking compute partnerships utilizing NVIDIA A100/H100 clusters or Google TPUs (v5e/v5p).