We're offering a great DPhil opportunity to look at comparing E. coli strain dynamics in gut in health and disease with ultimate aim of developing novel therapeutics. Opportunity to learn wet and dry lab skills including long-read metagenomics, biostastical + ex-vivo modelling. Happy to chat pls RT!

I often see instructions to cut word count in the same email with reviewers comments on additional paragraphs for the intro, methodology clarifications, additional papers to cite, limitations to acknowledge, "could the authors also comment on...."

I think positivity is with a denominator of all patients getting cultured, so implies that less is less - i.e. cultures no better targeted on real bacteraemia, so assuming patient mix did not change, implies that culture shortage led to less detection in absolute terms?

And we see such complex algorithms in our minds for responding to CRP levels, all of them unevidenced and often leading to more antibiotics.

Agree, although I think even with OVIVA and bone infection practice generally there is a high degree of pre-op planning - e.g. implications of retention of metalwork vs removal for treatment plan, adequacy of debridement possible or not possible and so on.

Cutaneous leishmaniasis is a major problem in Ethiopia & often requires admission for systemic therapy. Liquid nitrogen cryotherapy is an option but also has significant logistic barriers. We evaluated leveraging CO2 cryotherapy deployed for cervical cancer to decenteralise treatment for CL.

Vapes can bring their own problems, but hard to deny the importance of quitting tobacco, the difficulty of quitting, and that they seem to help.

The bioinformatic criteria that support clinical metagenomics as important as the wet lab workflow, and needs principled approach with derivation and validation to avoid overfitting.

If we say "any read counts" then lots of false +ve. If "only lots of reads" then lots of false -ve. Best compromise is ≥2 reads as threshold. But including number of reads in other samples in the same flowcell as an additional criteria helps exclude barcode crossover.

Detailed reflections: Serious technical limits applying metagenomics at scale. Lots of host DNA, and at scale need multiple samples per flow cell. So often trying to detect a few reads. And combining samples means crossover between samples, giving false positives. Bioinformatics is key.

Motivation was that using targeted PCR is limiting, want untargeted approach without relying on many primers. So metagenomics. We used workflow with ONT, 51% sensitive all viruses and 99.8% specific. Can improve sens to 70% by bioinformatics (more below) and to 83% excluding low level viruses.

Microbiology is coming out of the lab! Infusoria, a new exhibition by MMM Artist-in-Residence Anna Dumitriu, runs in the JR Hospital from 26 Sept–6 Nov 2026. New works inspired by Oxford research explore hospital sinks, wastewater, water microbiomes and antimicrobial resistance.