Biology subjects

Tzanakakis, A.

Publications and source records attributed to Tzanakakis, A..

3 recordsLinked to original sources

bioRxiv2026

Calibration-free compression brings Evo 2 to its full million-token context on a single GPU

Evo 2 is the largest openly available genomic foundation model, but its forty billion parameter configuration cannot be loaded onto a single 80 GB accelerator, placing genome-scale analysis beyond most laboratories. We present TurboQuant-Bio, an open toolkit that compresses Evo 2s weights and attention cache to four bits without calibration data, and serves both through fused kernels. Compression is near-lossless across perplexity spanning the tree of life, genomic classification, splice-site prediction, gene completion and clinically relevant variant-effect prediction. It brings Evo 2 40B onto one 80 GB GPU and Evo 2 7B to its full million-token context within a 40 GB memory budget, an eightfold gain in reachable context. We further show that the released chunked-prefill path is silently incorrect, returning plausible but uncorrelated likelihoods, and derive the block-wise continuation that repairs it: a complete 580-kilobase bacterial genome is now scored in one context in 22 minutes rather than 13.7 hours.

bioinformatics↗

bioRxiv2026

Fundamental limitations of genomic language models for realistic sequence generation

Large language models (LLMs) have shown remarkable success in natural language processing, prompting interest in their application to genomic sequence analysis. Genomic Language Models (gLMs) based on similar architectures offer a promising avenue for synthetic genome generation and characterization. However, their effectiveness for biological sequence modeling remains poorly characterized. We present a comprehensive evaluation of genomic language models that explicitly aim to generate entire synthetic genomes. We tested Evo 2 on diverse prokaryotic, eukaryotic and viral genomes, and megaDNA on bacteriophage genomes, and assessed performance across key biological features and organizational patterns. Our results reveal systematic failures in gLM-based genomic reconstruction. While the synthetic sequences captured local sequence statistics, they consistently failed to preserve long-range genomic organization, repeat and k-mer composition, transcription factor binding site architecture, and evolutionary constraints. Generated sequences exhibited violations of natural genomic patterns and models showed particular difficulty with repetitive elements. To assess the quality of genome generation, we trained a convolutional neural network that reliably distinguished synthetic from natural sequences, achieving AUROC values up to 0.97 in eukaryotes and 0.82 in prokaryotes, with classification accuracy increasing monotonically with genomic distance from the seed. These findings suggest fundamental limitations in current gLM architectures for capturing the long-range, hierarchical nature of genomic sequences. Our work highlights the need for specialized architectures that explicitly model evolutionary constraints rather than relying solely on statistical patterns, with important implications for computational biology applications requiring realistic sequence generation and for biosafety assessments that depend on the distinguishability of synthetic and natural genomic sequences.

synthetic biology↗

bioRxiv2026

Accelerating Inference in Genomic Foundation Models via Speculative Decoding

Genomic and protein foundation models (GFMs and PFMs) have demonstrated strong performance in learning the language of DNA and proteins, but their use in large-scale sequence generation is limited by the latency of autoregressive decoding. Because every token triggers a forward pass of a large Transformer, whose inference is relatively slow, long-sequence generation quickly becomes costly. In this work we adapt speculative decoding to a representative GFM: the DNA model DNAGPT and two representative PFMs: ProGen2 and ProtGPT2. We implement a probabilistic variant of speculative decoding, in which a lightweight draft model proposes short token spans and a larger target model verifies or corrects them in parallel, while preserving the target models sampling distribution. Across all three models we systematically study the effect of speculation window length, temperature, draft architecture and prompt length, and we benchmark tokens per second over multiple runs per configuration. Speculative decoding yields consistent speedups over standard key-value cached decoding, with maximum observed speedup reaching 100% increase, while average gains across models ranging between 20% and 40% (e.g. 1.2x-1.4x), without changing the underlying target model predictions. Our results show that speculative decoding is a practical and model-agnostic strategy for accelerating genomic and proteomic sequence generation without sacrificing prediction quality.

genomics↗