Matthew J Shepherd

FT comments section this morning - saying what everyone else is thinking, right?

Predicting protein-protein interactions (PPIs) at proteome scale can take months with co-folding models due to the massive all-vs-all comparisons required.

We are excited to announce FlashPPI, a contrastive learning framework that predicts proteome wide physical interfaces in minutes. 1/🧵

Epistasis and co-adaptation in bacterial genome evolution Nature Reviews Genetics - Bacterial genome evolution is shaped by epistasis, which can constrain or promote specific evolutionary paths. The authors review how integrating the effects of epistatic...

I’ve spent a lot of time over the past two years thinking about genome evolution and how recent research revealing complex gene-gene interactions has complicated traditional ‘gene-centric’ genomic approaches to evolution. So here is a review that consolidates these thoughts!

rdcu.be/e6pHY

Make no mistake this is a concerted and organised attempt to completely undermine science and expert opinion. This is Covid denialism on steroids and it’s on all of academia to see this for what it is and push back as strongly as possible

Priority effects inhibit the repeated evolution of phototrophy - npj Complexity npj Complexity - Priority effects inhibit the repeated evolution of phototrophy

New paper out! Here's a puzzle: phototrophy, the ability to use light for energy, is one of life's great innovations. It evolved early and transformed the biosphere. But it evolved 2x. Why not just once, why not more? Our work suggests the answer is priority effects.

www.nature.com/articles/s44...

This trip through Aotearoa New Zealand has provided one illustration after another about the value of heroic conservation efforts.

MENI is back! Join us in Dublin this August 2026 for our 3rd Meeting for Microbial Evolution in Ireland. We are delighted to have @rachelmwheatley.bsky.social @drrebeccajhall.bsky.social @jpjhall.bsky.social and @tweethinking.bsky.social join us as keynote speakers this year. miniurl.com/MENI

Cancer therapies exploit DNA repair defects to kill tumours. We asked whether the same logic could constrain antibiotic resistance evolution...

www.biorxiv.org/content/10.6...

Thank you!!!

It has been a pleasure to work on this huge study with @brockhurstlab.bsky.social, Niamh Harrington, @taoranfu.bsky.social, @jofoth.bsky.social, @scottishwormboy.bsky.social, Claudia Igler, Dylan Childs, Kenny Cagney, Anastasia Kotarra and Beth Grimsey (21/21)

Additionally, we identify fitness costs that shaped treatment-driven resistance dynamics in some patients, that could be leveraged to improve treatment, but that fitness costs posed little barrier to long-term persistence of resistance in other patients (20/21)

Together, our study outlines multiple ways AMR diagnosis and treatments could be improved by integrating knowledge of evolutionary processes - Detecting pre-existing resistance is highly important, and could predict how long treatment is likely to be successful for (19/21)

Our findings highlight that even for patients with the same infection type, and treatment with the same antimicrobial, there are multiple, varied and complex ecological and evolutionary mechanisms by which resistance can emerge (18/21)

This suggests an ecological alteration between distinct subpopulations, and that maintenance of this diversity within the infection may explain resistance oscillations. (17/21)

So why is costly resistance accompanied by cyclical dynamics in oscillatory trajectories, but not in monotonic ones? We found that higher genetic distances between pairs of samples with different treatment phases, versus the same phases, for oscillatory trajectories only (16/21)

Monotonic trajectories were significantly associated with spontaneous mutations, and despite frequently being accompanied by fitness costs there was low negative selection during off phases - evidence that costly resistance mutations can persist in vivo, even when treatment was withdrawn (15/21)

Average changes in resistance allele frequency matched the phenotypic dynamics we observed. In particular, for oscillatory trajectories, there was strong positive selection on resistance alleles during on phases, and strong negative selection during off phases (14/21)

We then tested whether the different trajectories we had observed corresponded to different underlying allele frequency dynamics for resistance-associated loci. We found these were predictive of trajectory in 77% of patients. (13/21)

The dominance of pre-existing resistance was additionally concerning, as it accelerated time taken for infections to become majority-resistant - with most of these cases gaining breakpoint resistance by day 14 of the year-long trial. (12/21)

Several patients had infections which displayed multiple of these mechanisms - highlighting that the picture of within-patient AMR evolution is complex even for a monotherapy of a single type of infection. (11/21)

Selection on pre-existing resistance present in the infection was the most common single evolutionary mechanism by which resistance emerged during treatment, followed by spontaneous mutations, and then immigration of a resistant strain into the infection site. (10/21)

Next, we combined this phenotypic dataset with whole genome sequencing of 4206 isolates, and all-isolate pools for each timepoint for each patient. Using this we were able to classify the evolutionary mechanism by which resistance emerged (9/21)

For oscillatory trajectories only, data significantly clustered by treatment dosing cycles, indicating treatment-driven switching between a high MIC / low growth rate, and a low MIC / high growth rate subpopulations - evidence treatment driving resistance-growth trade-off dynamics in-patient (8/21)

We then investigated whether these trajectories were driven by trade-offs between resistance and growth rate. We found a significant negative correlation between resistance and growth for monotonic and oscillatory trajectories, but not Stable ones. (7/21)

Additionally, we could group many patients into one of three 'MIC trajectories', based on how this phenotype changed over time during the trial - Stable (little/no change), Monotonic (a linear increase), and Oscillatory (MIC increasing and decreasing over time in line with treatment cycles) - (6/21)

After testing ciprofloxacin minimum inhibitory concentration (MIC) and growth kinetics for all isolates, we first found that MICs increased for the treatment group as the trial progressed, evidencing selection for increased resistance within-patient. (5/21)

All of these patients were sampled before and during a year-long phase-III clinical trial - ORBIT-3 - testing the efficacy of a ciprofloxacin-based inhalational therapy. The trial featured pulsed-treatment cycles, with drug dosed and withdrawn in 28 cycles. (4/21)

We set out to investigate in-depth how the process of treatment-induced resistance plays out in a large patient population, sampling 24,478 isolates of Pseudomonas aeruginosa from 180 people with bronchiectasis lung infections, a chronic inflammatory lung condition. (3/21)

Ecological and evolutionary mechanisms driving within-patient emergence of antimicrobial resistance - Nature Reviews Microbiology In this Review, Shepherd, Brockhurst and colleagues explore the clinical evidence in support of four major ecological and evolutionary mechanisms of within-patient antimicrobial resistance emergence i...

Whilst the molecular mechanisms and transmission of AMR are well defined, understanding of the evolutionary mechanisms by which resistance emerges, and of in situ fitness trade-off dynamics are limited by a lack of study scale. We reviewed this here: www.nature.com/articles/s41... (2/21)

🚨 New pre-print! 🚨 In the largest study of its kind to-date, we investigate the ecological and evolutionary mechanisms driving within-patient evolution of antimicrobial resistance (AMR). Read here:
www.biorxiv.org/content/10.6... , and follow along with this thread, discussing our findings (1/21)