←
Iain Johnston1,808 posts
Iain Johnston avatar

Iain Johnston

@mitomaths

Stochastic Biology Group: using maths, stats, and experiments to learn about biology and disease. We love organelles and oDNA, evolution, noise, and inference.

◷ Joined January 2013▧ 383 Photos and videos
Iain Johnston avatar
Iain Johnston@mitomaths·Sep 4

How does AMR evolution vary across pathogens? Regions? Patients? We used EvAM* to take a look. biorxiv.org/content/10.64898… Inferring evol pathways to multi-drug resistance as transition networks, we use the amazing CABBAGE dataset to learn AMR "roadmaps" across global ESKAPEEs... 1/

Post attachment
Iain Johnston avatar
Iain Johnston@mitomaths·Sep 4

... across several decades of clinical study. Would love to hear feedback, and on the aligned work on Klebsiella natural histories journals.plos.org/plosbiolog… and on *the perspective of EvAM in AMR evolution journals.asm.org/doi/full/10… 3/

Iain Johnston avatar
Iain Johnston@mitomaths·Sep 4

... finding striking similarities of evol dynamics across different pathogens, but also diffs predicted by geography and patient demographics. The networks (grey) are predictive, and perform well forecasting new resistome phenotypes (coloured dots) from sub-Saharan Africa... 2/

Post attachment
Iain Johnston avatar
Iain Johnston@mitomaths·Sep 4

... finding that the trained models do well at predicting multi-drug resistance states over several decades of clinical sampling. Related: Klebsiella-focussed natural history journals.plos.org/plosbiolog… and *EvAM-AMR perspective journals.asm.org/doi/full/10…

Iain Johnston avatar
Iain Johnston@mitomaths·Sep 4

... finding remarkable consistency across pathogen evolutionary "roadmaps", but some notable diffs which covary with region and patient demography. The inferred roadmaps also predict unseen and future resistance states, which we test with new data from sub-Saharan Africa... 2/

Post attachment
Iain Johnston avatar
Iain Johnston@mitomaths·Jul 12

A collection of summer mitochondria and chloroplasts, along with their environments 🇳🇴

Post attachmentPost attachmentPost attachmentPost attachment
Iain Johnston avatar
Iain Johnston@mitomaths·Jul 10

Mitochondria and chloroplasts are inherited through a "bottleneck". Smaller bottleneck ➡️ more powerful segregation of DNA damage. So why don't organisms just always use a bottleneck of size 1? And what's up with paternal leakage, DUI, etc? Take a look! doi.org/10.1093/gbe/evag161

Post attachment
Math Files avatar
⟳ Iain Johnston retweeted
Math Files@Math_files·Jun 26

Tupper’s formula is a formula that represents itself when one plots all the points in the plane that satisfy it.

Post attachment
Tim McNamara avatar
⟳ Iain Johnston retweeted
Tim McNamara@timClicks·Jun 26

Today in cursed code: a computer program that plays pong by modifying its own source code. (Each frame is a new compilation). github.com/uellenberg/Insert

0:17
ShorterLab avatar
⟳ Iain Johnston retweeted
ShorterLab@ShorterLab·Jun 10

The majority of axonal mitochondria in mammalian neurons lack mitochondrial DNA and do not produce ATP biorxiv.org/content/10.1101/…

Iain Johnston avatar
Iain Johnston@mitomaths·Jun 9

This sets up for a version of the Mk model that scales linearly in observations instead of exponentially in features, and can handle interactions, reversibility, coupling, missing data, and more. We think this could be quite a generally useful advance! 4/

Post attachment
Iain Johnston avatar
Iain Johnston@mitomaths·Jun 9

One can -- the classic Mk model in evol bio. But it struggles with many interacting features, because the possible state space becomes huge. But, though possible, many of these states are unlikely. So, basically, we prune them in a rational way. 3/

Post attachment
Iain Johnston avatar
Iain Johnston@mitomaths·Jun 9

If features influence each other (resistance to drug X makes resistance to drug Y more likely), that's a complication. If features can be lost as well as gained, that's another. If there are lots of them, that's another. Few methods can handle all that. 2/