New platform to strengthen our global defence against future pandemics - Novo Nordisk Fonden

Great news for our future!

Professor Samir Bhatt @sjbhatt.bsky.social
is joining a new initiative to provide free access to advanced AI tools for pathogen analysis to close gaps in global preparedness to enable faster responses to future outbreaks.
🎉 Samir!
novonordiskfonden.dk/en/news/new-...

Socioeconomic and temporal heterogeneity in SARS-CoV-2 exposure and disease in England from May 2020 to February 2023 Deprivation and ethnicity influenced COVID-19 outcomes, revealing health inequalities and vaccine effectiveness in the pandemic.

Read more about socioeconomic & temporal heterogeneity in SARS-CoV-2 exposure in England (May 2020 - Feb 2023) in this @science.org publication 👇
doi.org/10.1126/scia...

@imperialcollegeldn.bsky.social@imperialsph.bsky.social @thomrawson.bsky.social @sjbhatt.bsky.social @cm401.bsky.social

Socioeconomic and temporal heterogeneity in SARS-CoV-2 exposure and disease in England from May 2020 to February 2023 Deprivation and ethnicity influenced COVID-19 outcomes, revealing health inequalities and vaccine effectiveness in the pandemic.

We are excited to have a new paper published in Science Advances on Socioeconomic and temporal heterogeneity in SARS-CoV-2 exposure and disease in England — a great team effort led by @sjbhatt.bsky.social and Prof Neil Ferguson, in collaboration with the whole team.

www.science.org/doi/10.1126/...

Thoughts on how AI can help prepare for the next pandemic - Thanks for having led this work @mugkraemer.bsky.social Joseph L.-H. Tsui Serina Y. Chang, @sjbhatt.bsky.social - it is an honour to be part of it!

Artificial intelligence for modelling infectious disease epidemics Nature - This Perspective considers the application to infectious disease modelling of AI systems that combine machine learning, computational statistics, information retrieval and data science.

AI is poised to accelerate understanding in infectious diseases, but its value needs to be demonstrated through close collaboration between research, industry, society, and policy.

Paper free to read: rdcu.be/eaxEw

Summary here: www.ox.ac.uk/news/2025-02...

New @nature.com
How #AI can play a pivotal role in future pandemic preparedness and mitigation, with caveats
nature.com/articles/s41...
a privilege to join this global collaborative effort led by Moritz Kraemer and Samir Bhatt

Looking forward to digging into this. Co-authors include
@mghafari.bsky.social & @mugkraemer.bsky.social.
"Large-scale genomic surveillance reveals immunosuppression drives mutation dynamics in persistent SARS-CoV-2 infections"
www.medrxiv.org/content/10.1...

Alessandro Micheli, M\'elodie Monod, Samir Bhatt
Diffusion Models for Inverse Problems in the Exponential Family
https://arxiv.org/abs/2502.05994

The rising threat of deadly diseases jumping from animals to humans Zoonotic pathogens very likely caused the last pandemic. Can we get better at halting them before the next one?

The rising threat of deadly diseases jumping from animals to humans

www.ft.com/content/a714...

Ill be very curious to see how this new foundation model performs on different data, and will be pleasantly surprised if boosting is finally outperformed, its been king for a while....

Why do tree-based models still outperform deep learning on tabular data? While deep learning has enabled tremendous progress on text and image datasets, its superiority on tabular data is not clear. We contribute extensive benchmarks of standard and novel deep learning met...

Tabular data in contrast can be very heterogeneous and importantly, highly non-smooth, both of which can cause issues in deep neural networks. Tree methods tend to be be robust to uninformative features and orientations
arxiv.org/abs/2207.08815

The Intrinsic Dimension of Images and Its Impact on Learning It is widely believed that natural image data exhibits low-dimensional structure despite the high dimensionality of conventional pixel representations. This idea underlies a common intuition for the r...

Deep learning seems unreasonably effective. But many areas it succeeds in, the data have nice properties. For example, images, arxiv.org/abs/2104.08894 . Which might seem complex, but lie on low-dimensional manifold. Most images are locally connected, which is why spatial analysis works so well.

Accurate predictions on small data with a tabular foundation model - Nature Tabular Prior-data Fitted Network, a tabular foundation model, provides accurate predictions on small data and outperforms all previous methods on datasets with up to 10,000 samples by a wide margin.

An interesting paper out today www.nature.com/articles/s41...

As well known to many of us, tabular data is fiendishly hard, and whenever a new method comes out, boosting still beats it. Always, but this paper suggests deep learning is finally competitive.

Took us some time to get this out. Shows how one metric fits all can be dangerous. Thanks to all co-authors.

Ethnicity and anthropometric deficits in children: A cross-sectional analysis of national survey data from 18 countries in sub-Saharan Africa Child anthropometric deficits remain a major public health problem in Sub-Saharan Africa (SSA) and are a key target of the UN Sustainable Development Goals (SDGs). The SDGs recommend disaggregation of...

Stunting/wasting in children in Sub-Saharan Africa is a major public health concern. However, we show that identifying where the problem occurs can be difficult due to confounding with ethnicity.

journals.plos.org/globalpublic...

While this is all quite technical, the framework we present and the code is relatively simple, and opens new avenues to using renewal type models to estimate importation and transmission dynamics from observed data.

We also utilise the popular result of Barbour and Reinert to toggle a stochastic-deterministic switch to make computation faster. Using this framework we can calculate useful statistics like first passage times and times to extinction.

In this new work, we have tried to apply this framework to looking at importation dynamics. The long story short is, we extend our process to a marked Poisson process and are still able to numerically compute (via vectorised computation) the probability generating function.

Intrinsic randomness in epidemic modelling beyond statistical uncertainty - Communications Physics Intrinsic randomness is a critical source of uncertainty in infectious disease outbreaks. The authors show in a series of analytical results how this source of uncertainty can be better characterised.

In a second paper we derived the probability generating function of a time varying Crump-Mode-Jagers process and applied it to look at uncertainty (www.nature.com/articles/s42...).

Unifying incidence and prevalence under a time-varying general branching process - Journal of Mathematical Biology Renewal equations are a popular approach used in modelling the number of new infections, i.e., incidence, in an outbreak. We develop a stochastic model of an outbreak based on a time-varying variant o...

In a first paper we proved how the renewal equation used in infectious disease modelling arises from a Crump-Mode-Jagers (and, interestingly, also a Bellman-Harris) branching process (link.springer.com/article/10.1...).

Modelling the stochastic importation dynamics and establishment of novel pathogenic strains using a general branching processes framework The importation and subsequent establishment of novel pathogenic strains in a population is subject to a large degree of uncertainty due to the stocha…

New paper from Jacob and Frederik in the group. We have been interested in branching processes for a while now.

www.sciencedirect.com/science/arti...