Studying pathogens degrades BLAST-based pathogen identification

Abstract

As synthetic biology becomes increasingly capable and accessible, it is likewise increasingly critical to be able to make accurate biosecurity determinations regarding the pathogenicity or toxicity of particular nucleic acid or amino acid sequences. At present, this is typically done using the BLAST algorithm to determine the best match with sequences in the NCBI nucleic acid and protein databases. Neither BLAST nor any of the NCBI databases, however, are actually designed for biosafety determination. Critically, taxonomic errors or ambiguities in the NCBI nucleic acid and protein databases can also cause errors in BLAST-based taxonomic categorization. With heavily studied taxa and frequently used biotechnology tools, even low frequency taxonomic categorization issues can lead to high rates of errors in biosecurity decision-making. Here we focus on the implications for false positives, finding that BLAST against NCBI’s protein database will now incorrectly categorize a number of commonly used biotechnology tool sequences as the pathogens or toxins with which they have been used. Paradoxically, this implies that problems are expected to be most acute for the pathogens and toxins of highest interest and for the most widely used biotechnology tools. We thus conclude that biosecurity tools should shift away from BLAST against general purpose databases and towards new methods that are specifically tailored for biosafety purposes.

The whack-a-mole governance challenge for AI-enabled synthetic biology: literature review and emerging frameworks

“…The more accessible ( Wang and Zhang, 2019 ) and generally useful synbio potentially is regarded to be ( Sun et al, 2022 ), the less likely it is that prohibition will remain an effective tool. Decades-old bioinformatics resources originally developed to compare gene sequences, such as BLAST, have been re-used, with mixed results, as biosafety tools to identify pathogens ( Beal et al, 2023 ). Newer tools, such as machine learning-based topic models, enable spotting trends across a wide set of biosafety research publications ( Guan et al, 2022 ).…”

Ensemble Detection of DNA Engineering Signatures

“…Addressing this issue required weeks of SME effort to curate genomes for new target organisms and build exclusion lists for genomes that contained engineering. In addition, recent studies have highlighted the pitfalls of using general purpose sequence repositories such as NCBI to train screening systems . Consequently, new databases and methods are needed to aggregate and curate natural sequence data for the purposes of biosecurity.…”

“…This sequence analysis module is based on techniques from malware detection that were later adapted for pathogenic DNA screening as part of BBN’s FAST-NA tool . BGAF includes whole-genome assembly with Abyss and Prymetime (our custom hybrid assembly pipeline) taxonomic classification and read mapping with FAST-NA, and annotation with BLAST.…”

Beyond Biosecurity by Taxonomic Lists: Lessons, Challenges, and Opportunities

“…Other tools (eg, sequence databases) may have been created for other purposes and may need to be reconfigured or curated for use in biosecurity regimes. 27 Some tools needed by biosecurity regimes are not yet fully developed (eg, algorithms to identify harmful functions from primary DNA or protein sequences). When developing tools, it is vital to take into account the need to avoid unduly impeding biological research and development and to manage any information hazards generated.…”

“…Commercial providers of DNA synthesis have noted the challenges of implementing taxonomy-based regulations for sequence-based products. 26 , 27 Ascertaining whether a given order matches something to be controlled is difficult even for the most qualified companies. Taxonomic lists have also been used as a tool to narrow oversight of relevant biological research.…”