We propose the first unified adversarial attack benchmark for Genomic Foundation Models (GFMs), named GERM. Unlike existing GFM benchmarks, GERM offers the first comprehensive evaluation framework to systematically assess the vulnerability of GFMs to adversarial attacks. Methodologically, we evaluate the adversarial robustness of five state-of-the-art GFMs using four widely adopted attack algorithms and three defense strategies. Importantly, our benchmark provides an accessible and comprehensive framework to analyze GFM vulnerabilities with respect to model architecture, quantization schemes, and training datasets. Empirically, transformer-based models exhibit greater robustness to adversarial perturbations compared to HyenaDNA, highlighting the impact of architectural design on vulnerability. Moreover, adversarial attacks frequently target biologically significant genomic regions, suggesting that these models effectively capture meaningful sequence features.
Decoding the linguistic intricacies of the genome is a crucial problem in biology, and pre-trained foundational models such as DNABERT and Nucleotide Transformer have made significant strides in this area. Existing works have largely hinged on k-mer, fixed-length permutations of A, T, C, and G, as the token of the genome language due to its simplicity. However, we argue that the computation and sample inefficiencies introduced by k-mer tokenization are primary obstacles in developing large genome foundational models. We provide conceptual and empirical insights into genome tokenization, building on which we propose to replace k-mer tokenization with Byte Pair Encoding (BPE), a statistics-based data compression algorithm that constructs tokens by iteratively merging the most frequent co-occurring genome segment in the corpus. We demonstrate that BPE not only overcomes the limitations of k-mer tokenization but also benefits from the computational efficiency of non-overlapping tokenization. Based on these insights, we introduce DNABERT-2, a refined genome foundation model that adapts an efficient tokenizer and employs multiple strategies to overcome input length constraints, reduce time and memory expenditure, and enhance model capability. Furthermore, we identify the absence of a comprehensive and standardized benchmark for genome understanding as another significant impediment to fair comparative analysis. In response, we propose the Genome Understanding Evaluation (GUE), a comprehensive multi-species genome classification dataset that amalgamates $36$ distinct datasets across $9$ tasks, with input lengths ranging from $70$ to $10000$. Through comprehensive experiments on the GUE benchmark, we demonstrate that DNABERT-2 achieves comparable performance to the state-of-the-art model with $21 \times$ fewer parameters and approximately $92 \times$ less GPU time in pre-training.
DNA sequencing is revolutionising the field of medicine. DNA sequencers, the machines which perform DNA sequencing, have evolved from the size of a fridge to that of a mobile phone over the last two decades. The cost of sequencing a human genome also has reduced from billions of dollars to hundreds of dollars. Despite these improvements, DNA sequencers output hundreds or thousands of gigabytes of data that must be analysed on computers to discover meaningful information with biological implications. Unfortunately, the analysis techniques have not kept the pace with rapidly improving sequencing technologies. Consequently, even today, the process of DNA analysis is performed on high-performance computers, just as it was a couple of decades ago. Such high-performance computers are not portable. Consequently, the full utility of an ultra-portable sequencer for sequencing in-the-field or at the point-of-care is limited by the lack of portable lightweight analytic techniques. This thesis proposes computer architecture-aware optimisation of DNA analysis software. DNA analysis software is inevitably convoluted due to the complexity associated with biological data. Modern computer architectures are also complex. Performing architecture-aware optimisations requires the synergistic use of knowledge from both domains, (i.e, DNA sequence analysis and computer architecture). This thesis aims to draw the two domains together. In this thesis, gold-standard DNA sequence analysis workflows are systematically examined for algorithmic components that cause performance bottlenecks. Identified bottlenecks are resolved through architecture-aware optimisations at different levels, i.e., memory, cache, register and processor. The optimised software tools are used in complete end-to-end analysis workflows and their efficacy is demonstrated by running on prototypical embedded systems.