Thomas Doherty | arXiv Science

arXiv2024

Machine learning augmented diagnostic testing to identify sources of variability in test performance

Diagnostic tests that can detect pre-clinical or sub-clinical infection, are one of the most powerful tools in our armoury of weapons to control infectious diseases. Considerable effort has been paid to improving diagnostic testing for human, plant and animal diseases, including strategies for targeting the use of diagnostic tests towards individuals who are more likely to be infected. We use machine learning to assess the surrounding risk landscape under which a diagnostic test is applied to augment its interpretation. We develop this to predict the occurrence of bovine tuberculosis incidents in cattle herds, exploiting the availability of exceptionally detailed testing records. We show that, without compromising test specificity, test sensitivity can be improved so that the proportion of infected herds detected improves by over 5 percentage points, or 240 additional infected herds detected in one year beyond those detected by the skin test alone. We also use feature importance testing for assessing the weighting of risk factors. While many factors are associated with increased risk of incidents, of note are several factors that suggest that in some herds there is a higher risk of infection going undetected.

cs.LG↗

arXiv2022

Modelling plausible scenarios for the Omicron SARS-CoV-2 variant from early-stage surveillance

We used a spatially explicit agent-based model of SARS-CoV-2 transmission combined with spatially fine-grained COVID-19 observation data from Public Health Scotland to investigate the initial rise of the Omicron (BA.1) variant of concern. We evaluated plausible scenarios for transmission rate advantage and vaccine immune escape relative to the Delta variant based on the data that would have been available at that time. We also explored possible outcomes of different levels of imposed non-pharmaceutical intervention. The initial results of these scenarios were used to inform the Scottish Government in the early outbreak stages of the Omicron variant. Using the model with parameters fit over the Delta variant epidemic, some initial assumptions about Omicron transmission rate advantage and vaccine escape, and a simple growth rate fitting procedure, we were able to capture the initial outbreak dynamics for Omicron. We found that the modelled dynamics hold up to retrospective scrutiny. The modelled imposition of extra non-pharmaceutical interventions planned by the Scottish Government at the time would likely have little effect in light of the transmission rate advantage held by the Omicron variant and the fact that the planned interventions would have occurred too late in the outbreak's trajectory. Finally, we found that any assumptions made about the projected distribution of vaccines in the model population had little bearing on the outcome, in terms of outbreak size and timing. Instead, it was the landscape of prior immunity that was most important.

q-bio.PE↗