@andimscience

Quantitative immunology & biological physics. Q-Immuno Lab PI @iit_ucl & @ucl_ipls past: Lewis-Sigler Fellow @Princeton, PhD @ENS_ULM Also @qimmuno.bsky.social

 London, UK
qimmuno.com
Joined February 2014
  • Tweets680
  • Following881
  • Followers1.2K
24Media
When comparing clusterings, what makes a better solution? In a new preprint, I describe homogeneity–parsimony curves as a means to quantify the underlying trade-off. Think of it as an ROC curve, but for clustering validation. arxiv.org/abs/2607.20799

External Clustering Validation by the Homogeneity-Parsimony Trade-off

Scalar metrics are often used to evaluate clusterings against known classes, but they can obscure a fundamental trade-off: clusterings should be informative about class labels while avoiding...

arxiv.org
3
7
760
One month until London Q-Immuno Day 2026! 🧬💻 Join us on May 8th for interdisciplinary discussions at the intersection of theory, computation and immunology. 🥳 More details and free registration at: qimmuno.com/ldnday/
2
7
815
Andreas Tiffeau-Mayer retweeted
Malcolm Sim@mjwsim
Mar 6
The Sim lab has been hard at work and we’re proud to present our first paper. We developed new methods to study peptide:KIR interactions. Congratulations to star postdoc Tanusya Murali who led the project. biorxiv.org/content/10.64898…
4
4
18
1,063
How different are self and nonself? This is a central question in immunology. In our latest work just published in @PRX_Life we demonstrate that at the peptide-level statistical differences between host and pathogen proteomes are minor
PRX Life@PRX_Life
Mar 4
A statistical physics framework that models peptidomes across species shows that self and nonself peptides are nearly one and the same, implying that the immune system benefits by targeting antigens near those represented in the organism’s own proteome. go.aps.org/40tAuIl
A plot depicts the divergence between peptides from different proteomes. The x-axis shows the Kullback-Leibler divergences between peptide distributions of different pathogen proteomes relative to human host peptides, and the y-axis shows the same divergences relative to uniform distribution peptides. Various colors of curved and angled lines represent human, mouse, chicken, zebrafish, human virome, P. falciparum, tuberculosis, listeria, and StrepA proteomes. The lines connect a first moment (amino acid frequencies), second moment (frequency covariance), and third moment (a third-order correlation measure), as well as a two-point (distance-specific pairwise covariation) — depicted with a circle, square, triangle, and diamond shape, respectively. A cluster of proteomes on the left side of the plot is magnified with an inset to show small differences between them.

ALT A plot depicts the divergence between peptides from different proteomes. The x-axis shows the Kullback-Leibler divergences between peptide distributions of different pathogen proteomes relative to human host peptides, and the y-axis shows the same divergences relative to uniform distribution peptides. Various colors of curved and angled lines represent human, mouse, chicken, zebrafish, human virome, P. falciparum, tuberculosis, listeria, and StrepA proteomes. The lines connect a first moment (amino acid frequencies), second moment (frequency covariance), and third moment (a third-order correlation measure), as well as a two-point (distance-specific pairwise covariation) — depicted with a circle, square, triangle, and diamond shape, respectively. A cluster of proteomes on the left side of the plot is magnified with an inset to show small differences between them.

1
4
1
18
1,875
more replies
An important implication for cancer treatment is that neoantigens that are just one mutation from self are not strange antigen at all, as sometimes thought.
1
4
90
More broadly, these findings might also help inform emerging protein design approaches to reduce unwanted immunogenicity
3
84
Excited to share the latest preprint from the lab "Fluctuating environments are sufficient to drive substantial variability in species abundance across locations" now up on arXiv arxiv.org/abs/2603.02403 🎉
1
7
36
1,837
more replies
Our work on these questions has been motivated by our experimental findings uncovering substantial variation in the human immune repertoire across blood and tonsils. doi.org/10.1016/j.immuni.202…
1
77
Watch out for future work in which we ask: how much spatial variability in the immune repertoire can be explained by localised clonal expansions?
66
If you're at @britsocimm Congress, come catch three talks today with updates about our research!
1
1
110
more replies
Later on I will give an invited talk in the session "AI driven insights into immune responses: vaccines, infections, and disease" laying out a vision for combining AI with TCRseq to study human immunity
1
96
Looking forward to connecting with many of you!
85
Load more