Biology subjects

Clifton, L.

Publications and source records attributed to Clifton, L..

3 recordsLinked to original sources

bioRxiv2026

MedMisBench: Measuring Epistemic Resilience of LLMs Under Misleading Medical Context

Large language models (LLMs) now reach expert-level scores on medical licensing exams, encouraging the assumption that high scores imply safe medical judgment while patients increasingly use them for health advice. We show this assumption is fragile: when misleading context is injected into questions that LLMs originally answer correctly, they abandon the correct answer. We call the ability to maintain correct judgment under adversarial context epistemic resilience, and introduce MedMisBench to measure it. MedMisBench contains 10,932 medical question items and 48,889 misleading context-option pairs spanning medical reasoning, agentic capability, and patient-journey evaluation. Across 11 model configurations, mean accuracy falls from 71.1% on original questions to 38.0% under focused misleading context, with 51.5% attack success. The most damaging injections are formal, rule-like fabrications: authority-framed falsehoods reach 69.5% attack success and exception-poisoning claims reach 64.1%. A 14-member clinical panel from 7 countries identified serious potential harm in 38.2% of reviewed cases. MedMisBench exposes a structural blind spot in LLM evaluation in medical settings: existing benchmarks measure what models know, but not whether they preserve correct medical judgment under misleading context.1

bioengineering↗

bioRxiv2026

Early plasma proteomic alterations precede amyloidosis diagnosis, reflecting cardiac and immune dysregulation

Systemic amyloidosis is typically diagnosed only after irreversible organ damage has occurred, limiting the effectiveness of available therapies. Whether the disease is preceded by detectable molecular changes long before clinical presentation has remained unclear. Here, we leveraged population-scale plasma proteomics and longitudinal follow-up from the UK Biobank to investigate early circulating protein signatures associated with future diagnosis of amyloidosis. Among approximately 53,000 participants with proteomic profiling, we identified 61 individuals who developed amyloidosis up to 14 years after protein assessment. Differential expression and correlation analyses identified a seven-protein panel, including MYL3, MYBPC1, NT-proBNP, NPPB, FCRLB, IGFBP1, and FABP1, consistent with early cardiac stress and immune dysregulation. Time-to-event modelling demonstrated robust stratification of amyloidosis risk and timing. Importantly, a parsimonious subset of these proteins retained strong predictive performance, indicating that a reduced set of biologically informative markers is sufficient for risk stratification. Furthermore, these proteomic signals were not explained by pre-existing cardiac disease, clonal haematopoiesis, or related plasma cell disorders, indicating that they capture disease-specific biological processes preceding clinical diagnosis. Together, these findings show that amyloidosis is preceded by persistent plasma proteomic alterations, providing a framework for early risk stratification and insight into the preclinical biology of this under-recognised disease.

bioinformatics↗

bioRxiv2022

Metabolism is correlative not causative for age-related auditory decline in an insect model

Aging is due to a complex decline of multiple biological processes. Some of the causes include oxidative damage, mitochondrial and proteostatic dysfunction, and DNA damage. The result is that as biological systems age their performance deteriorates. This age-related decline is well quantified, and experienced, for human hearing and is presumed to be due to a decrease in the ears metabolism - specifically a decrease in ability to maintain an electrochemical gradient, the endocochlear potential. However, direct measurements of metabolism across a lifespan in an auditory system are lacking. Even if metabolism does decrease with age, the question remains is it a cause of age-related auditory decline or simply correlative? All auditory systems across the animal kingdom share functional principles including ion pumping cells, auditory receptors, spiking auditory nerves and multiple supporting cells. Therefore, we used an insect, the desert locust, Schistocerca gregaria, as a physiologically versatile model to understand how cellular metabolism correlates with age and impacts on age-related auditory decline. We found that although metabolism correlates with age-related auditory decline it is not causative.

neuroscience↗