Biology subjects

Isaac, A.

Publications and source records attributed to Isaac, A..

5 recordsLinked to original sources

bioRxiv2026

An Expert-Curated Benchmark for Frontier and Local LLMs Querying a Biomedical Knowledge Graph

We combine GeneNetwork's 1TB structured database with its metadata as an RDF graph and an associated literature corpus. We benchmark frontier and local LLMs querying it through four architectures: document, graph and agentic retrieval-augmented generation (RAG), and a combination of the three. We evaluate them on 40 domain specific expert-curated questions with a new metric: the satisfaction frequency psi (psi, i.e, the share of questions answered with precision and recall of at least 70% or an f1 of at least 60%). psi ranks systems the same way precision, recall and f1 -- a machine learning metric built from the harmonic mean of precision and recall -- but is easier to read and assess for humans. System architecture had a large effect: with full use of Claude Opus 4.8 the plain document RAG of our earlier GNQA system had a psi of 36% against 86% for our new combined tools. Delegating most of the work from Claude Opus 4.8 to a local model halved query costs while still reaching a psi of 59%.

bioinformatics↗

bioRxiv2025

Narrowed host ranges do not constrain future host range expansion in RNA phage phi6

RNA viruses frequently shift between infecting different hosts and emerge on novel hosts. Part of this evolutionary process can involve specialization, when viruses adapt to increase their fitness on a particular host, often at the expense of their ability to infect other hosts. This trajectory has the potential to lead to extreme host narrowing which excludes all other previously accessible hosts. The consequences of the genetic architecture of host specialization on a viruss future evolutionary potential are understudied. In this study, we explored the ability of extreme specialists to re-expand their host ranges, particularly the ease and mutational mechanisms by which they might accomplish this such as reversion of host-range-narrowing mutations or mutations at other sites. Using previously evolved specialized strains of model dsRNA bacteriophage phi6 which had lost their ability to infect some hosts, we challenged specialists to adapt to their ancestral and other challenge hosts and identified resulting mutations. We found that these specialists readily re-gained their broader host ranges, at rates comparable to previously observed emergence events in phi6, indicating a lack of potential constraint from their mutational backgrounds due to epistasis. While some viral strains achieved host range re-expansion by reversing the original host-narrowing mutations gained during specialization, others used secondary mutations which were found to be parallel mutations previously associated with phi6 entry into those host species. This study contributes to our understanding of the evolutionary dynamics of host shifting in RNA viruses and their strategies to re-expand following specialization, which is relevant to spillback events and recurring host shifts that are observed in nature.

evolutionary biology↗

bioRxiv2024

Creating a biomedical knowledge base by addressing GPT inaccurate responses and benchmarking context

We created GNQA, a generative pre-trained transformer (GPT) knowledge base driven by a performant retrieval augmented generation (RAG) with a focus on aging, dementia, Alzheimers and diabetes. We uploaded a corpus of three thousand peer reviewed publications on these topics into the RAG. To address concerns about inaccurate responses and GPT hallucinations, we implemented a context provenance tracking mechanism that enables researchers to validate responses against the original material and to get references to the original papers. To assess the effectiveness of contextual information we collected evaluations and feedback from both domain expert users and citizen scientists on the relevance of GPT responses. A key innovation of our study is automated evaluation by way of a RAG assessment system (RAGAS). RAGAS combines human expert assessment with AI-driven evaluation to measure the effectiveness of RAG systems. When evaluating the responses to their questions, human respondents give a "thumbs-up" 76% of the time. Meanwhile, RAGAS scores 90% on answer relevance on questions posed by experts. And when GPT-generates questions, RAGAS scores 74% on answer relevance. With RAGAS we created a benchmark that can be used to continuously assess the performance of our knowledge base. Full GNQA functionality is embedded in the free GeneNetwork.org web service, an open-source system containing over 25 years of experimental data on model organisms and human. The code developed for this study is published under a free and open-source software license at https://git.genenetwork.org/gn-ai/tree/README.md.

scientific communication and education↗

bioRxiv2020

Quorum Sensing Regulates 'swim-or-stick' Lifestyle in the Phycosphere

Originality-significance statementMotility and biofilm formation are processes regulated by quorum sensing (QS) in bacteria. Both functions are believed to play an important role in interactions between bacteria and phytoplankton. Here, we show that two bacterial symbionts from the microbial community associated with a ubiquitous diatom switch their motile lifestyle to attached cells while an opportunist bacterium from the same community is incapable of attachment, despite possessing the genetic machinery to do so. Further work indicated that the opportunist lacks QS signal synthases while the symbionts produce three QS signals, one of which is mainly responsible for regulating symbiont colonization of the diatom microenvironment. These findings suggest that QS regulates colonization of diatom surfaces and further work on these model systems will inform our understanding of particle aggregation and bacterial attachment to marine snow and how these processes influence the global carbon cycle. SummaryInteractions between phytoplankton and bacteria play major roles in global biogeochemical cycles and oceanic nutrient fluxes. These interactions occur in the microenvironment surrounding phytoplankton cells, known as the phycosphere. Bacteria in the phycosphere use either chemotaxis or attachment to benefit from algal excretions. Both processes are regulated by quorum sensing (QS), a cell-cell signaling mechanism that uses small infochemicals to coordinate bacterial gene expression. However, the role of QS in regulating bacterial attachment in the phycosphere is not clear. Here, we isolated a Sulfitobacter pseudonitzschiae F5 and a Phaeobacter sp. F10 belonging to the marine Roseobacter group and an Alteromonas macleodii F12 belonging to Alteromonadaceae, from the microbial community of the ubiquitous diatom Asterionellopsis glacialis. We show that only the Roseobacter group isolates (diatom symbionts) can attach to diatom transparent exopolymeric particles. Despite all three bacteria possessing genes involved in motility, chemotaxis, and attachment, only S. pseudonitzschiae F5 and Phaeobacter sp. F10 possessed complete QS systems and could synthesize QS signals. Using UHPLC-MS/MS, we identified three QS molecules produced by both bacteria of which only 3-oxo-C16:1-HSL strongly inhibited bacterial motility and stimulated attachment in the phycosphere. These findings suggest that QS signals enable colonization of the phycosphere by algal symbionts.

ecology↗

bioRxiv2020

Diatom Modulation of Microbial Consortia Through Use of Two Unique Secondary Metabolites

Unicellular eukaryotic phytoplankton, such as diatoms, rely on microbial communities for survival despite lacking specialized compartments to house microbiomes (e.g., animal gut). Microbial communities have been widely shown to benefit from diatom excretions that accumulate within the microenvironment surrounding phytoplankton cells, known as the phycosphere. However, mechanisms that enable diatoms and other unicellular eukaryotes to nurture specific microbiomes by fostering beneficial bacteria and repelling harmful ones are mostly unknown. We hypothesized that diatom exudates may attune microbial communities and employed an integrated multi-omics approach using the ubiquitous diatom Asterionellopsis glacialis to reveal how it modulates its naturally associated bacteria. We show that A. glacialis reprograms its transcriptional and metabolic profiles in response to bacteria to secrete a suite of central metabolites and two unusual secondary metabolites, rosmarinic acid and azelaic acid. While central metabolites are utilized by potential bacterial symbionts and opportunists alike, rosmarinic acid promotes attachment of beneficial bacteria to the diatom and simultaneously suppresses the attachment of opportunists. Similarly, azelaic acid enhances growth of beneficial bacteria, while simultaneously inhibiting growth of opportunistic ones. We further show that the bacterial response to azelaic acid is widespread in the worlds oceans and taxonomically restricted to a handful of bacterial genera. Our results demonstrate the innate ability of an important unicellular eukaryotic group to modulate their microbial consortia, similar to higher eukaryotes, using unique secondary metabolites that regulate bacterial growth and behavior inversely in different bacterial populations.

microbiology↗