WWTFVienna Research Radar Subscribe
Cover: © Philipp Velicky, Core Facility Imaging, MedUni Wien

Issue #12 · 29 October 2026

Marrow, Moonlight, and Machines

The immune system's home guard gets a new supply line

Macrophages are the immune system's resident caretakers: they clear debris, fight infection and, inside tumours, often end up helping the cancer along. Where they come from has long been told in two parts. Before birth, they arise from progenitors in the yolk sac. In adults, new ones are thought to come from monocytes circulating in the blood.

Two companion papers in Nature Immunology, from a team at the Medical University of Vienna led by Shweta Tikoo, Rohit Jain and Wolfgang Weninger, with partners in Sydney and at the Weizmann Institute of Science, revise both parts. In adult mouse bone marrow, they identify a rare progenitor, EMPROR, that makes macrophages without passing through the monocyte route. It supplies macrophages around blood vessels and inside tumours, and its genetic profile sits closest to the embryonic progenitors, as if an early-life programme had carried on into adulthood. Searching for its embryonic counterpart, the team found a blood-forming site that had not been mapped: tissue at the interface between mother and embryo, developing alongside the yolk sac but separate from it. Early human placental tissue, examined with Jürgen Pollheimer's group at the same university, carries CD4-marked immature red blood cells, a first sign that the site matters in humans too.

Most of the evidence comes from mice, and the human part is a hint, not yet a proof. The practical consequence is already visible. Tumour-associated macrophages are a major target in cancer research, and strategies that stop monocytes from entering tissue may leave this second route open.

The proposal behind WWTF project LS18-013, funded in 2019, already described a marrow precursor unrelated to monocytes and resembling yolk-sac progenitors, and called it a challenge to a classical paradigm of blood-cell formation. A preprint followed in 2021. Making the case took until this autumn, and two papers rather than one.

→ Tikoo, Jain, Martinez et al., Nature Immunology, 2026, → Jain, Tikoo, Márquez-Burgoa et al., Nature Immunology, 2026

Under the moon, a clockwork orange keeps the worm's time

The marine bristle worm Platynereis dumerilii keeps a daily clock that runs at about 24 hours or about 24.8, the length of a lunar day, depending on the moon's phase. In a preprint, Federico Scaramuzza ( Max Perutz Labs Vienna ), Kristin Tessmar-Raible and colleagues at the University of Vienna, with Hanspeter Herzel in Berlin, find out why. Among the genes that follow the moon most closely is clockwork orange, or short cwo, a clock gene first described in fruit flies and related to Dec1 and Dec2 in mammals. A computer model shows it can explain the shift; worms with the gene edited out lose it. Corals and vertebrates feel the moon in their daily timing too.

→ Scaramuzza, Andreatta, Mat et al., bioRxiv, 2026 (preprint)

The one number AI still gets wrong about magnesium

Magnesium ions matter in countless biological processes, yet classical simulations cannot capture their structure, thermodynamics and kinetics at the same time. Sebastian Falkner at Universität Augsburg and University of Vienna, with Pablo Montero de Hijes and Christoph Dellago in Vienna and Nadine Schwierz in Augsburg, tests neural network potentials trained on quantum-chemical data in the Journal of Chemical Theory and Computation. They reproduce the shell of water around the ion, its diffusion, and even the rare moment a water molecule swaps out, within an order of magnitude of experiment. The one figure they miss is the solvation energy, the energy released when water closes around the ion. It is set by electrical forces from far away, and the models are built to look only close by.

→ Falkner, Montero de Hijes, Dellago, Schwierz, Journal of Chemical Theory and Computation, 2026

Seeding doubt, the weapon of choice under authoritarian rule

Research on false information mostly studies Western democracies and hunts for bots. Hossein Kermani at the University of Vienna and colleagues hand-coded 36,255 popular tweets from 4,807 accounts posted during the first months of Iran’s Women, Life, Freedom movement. Writing in Democratization, they find outright fabrication giving way to accusations of fabrication and to claims nobody could check. Where independent verification is all but impossible, unverifiability itself becomes a tool; coordinated bot campaigns played only a limited part. Detection systems that scan text for falsehoods would miss most of it. Part of WWTF project ICT23-002.

→ Kermani, Oudlajani, Yarahmadi et al., Democratization, 2026

This fortnight's Crossref data brought forward a short paper from Emanuel Sallinger 's group at the TU Wien, credited, among other WWTF grants, to VRG18-013. That Vienna Research Group was set up in 2020 to make machines reason at scale over knowledge graphs, networks of billions of linked facts. Its latest work turns from how AI systems reason to how they should behave. Alessandro Pesare and Tommaso Dolci, with Katja Hose and Sallinger, ask where, in a system of cooperating AI agents, values such as privacy, fairness and pluralism should live. Most safeguards check the final output, after biased reasoning or leaked data has already passed from agent to agent.

The paper, presented this month at a workshop on agentic AI in software development in Munich, builds them into the architecture instead. Domain agents never see each other's data. Deliberation runs without a central orchestrator that could quietly favour one view. Guard agents review a plan before anything is executed and hand hard cases to a human. The paper cites the Digital Humanism textbook, and reads like its principles turned into design patterns.

"I'm not a problem to be solved"

Sign language gloves, a favourite of labs and the press, promise to turn hand movements into speech. At TU Wien, Oliver Suchanek, who is Deaf, set out with Katta Spiel and Vanessa Schügerl of Heriot-Watt University to test a commercial one, with eleven Deaf peers. Even signs the glove had been trained on were recognised less than half the time. In a supermarket, GOOD MORNING came out as MORNING INTESTINE. Then they stopped. Gloves like this, they write, get funded, published and celebrated, while the people they are built for are left explaining, yet again, why they turn down help they never asked for. Its last sentence, in the first author's own voice, is its title: I'm not a problem to be solved.

Every image at the top of this newsletter so far has come from a Vienna lab. Many more never leave the place they were made. With ReFrame, the Science Image Award Vienna 2026, WWTF wants to bring the best of them into public view. Researchers in Vienna and the wider metropolitan region, Institute of Science and Technology Austria and International Institute for Applied Systems Analysis (IIASA) included, can submit a series of three to ten images from a single research project. Every discipline and every career stage is welcome, and so are data managers, research librarians, lab managers and science communicators. Microscopy counts, and so do fieldwork, archives, interviews and data. Generated or composite images do not. An independent jury awards EUR 3,000 to the winning series and EUR 1,000 each to two runners-up, with winners announced in early 2027. The deadline is 30 November 2026, 2 pm.

→ Suchanek, Schuegerl, Spiel, ASSETS '26, 2026

Subscribe on LinkedIn Browse all issues