EEB Daily

Vol. 1, No. 37Sunday, November 22, 2026Bryan, Texas
One paper worth the coffee. One paper from beyond the fence. One story from the wider world.

The lead · Population genomics: inferring inbreeding from sparse sequence data

Cheap genotypes still reveal inbreeding

Sparse genetic data can recover overall inbreeding even when the age of shared ancestry remains difficult to resolve.

When two copies of a DNA segment descend from the same ancestor, the segment is homozygous by descent (HBD). Long HBD tracts usually trace recent shared ancestors; short tracts can reflect more distant ancestry. That distinction matters when breeding programs or conservation managers need to measure inbreeding without paying for deep sequencing of every individual. Forneris and colleagues asked how reliably researchers can recover HBD from three cheaper genomic assays: small SNP panels, reduced-representation sequencing, and low-coverage whole-genome sequencing. They began with high-coverage genomes from 132 Holstein cattle, then computationally thinned the data to mimic each assay.

Using models that assign genomic stretches to multiple HBD length classes, the team recovered overall inbreeding coefficients with SNP panels of 6,000 markers, reduced-representation datasets of 15,000 SNPs, and sequencing depth as low as 0.1 times. The analyses also supported genome-wide and locus-specific estimates, but only the longest HBD tracts could be identified dependably. Tests with Nordic Red cattle and Labrador retrievers broadly confirmed the pattern using real economical data. The lesson is not that all ancestry can be reconstructed cheaply: model choice and the decision about which HBD classes to trust remain critical, particularly for distinguishing old from recent inbreeding.

Where to pushA useful follow-up would compare these thresholds in populations with lower diversity, strong population structure, or recent admixture. In small conservation populations, long HBD tracts might be common, but relatedness and marker ascertainment could differ from dairy cattle. Which sampling strategy gives the least biased prediction of deleterious recessive homozygosity per dollar spent?

Beyond EEB · A paper from outside evolution and ecology

Soft synthetic surfaces teach T cells to persist

Yang Y. et al., Nature Communications (2026).

Immune cells pay attention not only to molecular signals but also to the mechanical behavior of the surface delivering them. Researchers built synthetic antigen-presenting cells that could relax after deformation at different rates while keeping stiffness and antigen density comparable. The faster-relaxing surfaces strengthened T-cell receptor and mTOR signaling and shifted expanded CD8 T cells toward metabolic states associated with better persistence and memory-like properties. In experimental melanoma and ovarian cancer models, the resulting engineered T cells survived longer and showed more durable antitumor activity. This is a manufacturing strategy demonstrated in laboratory and animal models, not evidence that it improves outcomes in patients.

Beyond biology · A result from outside biology

An AI team designs crystals chemists can actually make

Inizan T. J. et al., Nature Computational Science (20 Nov 2026).

Metal-organic frameworks are crystalline scaffolds built from metal nodes and organic linkers; their pores can capture carbon dioxide or harvest water. A team combined specialized AI agents into MOFGen: language models suggested chemical ingredients, a diffusion model generated atomic structures, physics-based tools checked their geometry, and synthesis predictors screened which designs might be made. The system explored hundreds of thousands of candidate frameworks, and chemists synthesized four AI-suggested materials. That physical validation matters because a plausible crystal on a computer is not necessarily manufacturable. The authors note that experiments still relied on human intervention and tested only a narrow slice of the computational designs.