Issue #7 – What the Signal Is Worth · Vienna Research Radar

Spotlight

A urine test that catches kidney cancer before it declares itself

Clear cell renal cell carcinoma is the most common kidney cancer and the deadliest, and it has a habit of staying quiet. There is no screening test for it. Most cases surface by accident, spotted on a scan ordered for something else, or late, once the tumour causes pain or blood in the urine. By then a third of patients already have advanced disease. The organ we associate with filtering waste has offered no reliable early signal of its own most dangerous illness.

A study in EMBO Molecular Medicine, led by Gustav Jonsson, Tiago Oliveira and Maura Hofmann in Josef Penninger 's group at Institute of Molecular Biotechnology Austrian Academy of Sciences, with clinical colleagues at the Medical University of Vienna, went looking for that signal where it should logically appear: in urine. The kidney filters the blood, so a tumour sitting in that filtration system has every opportunity to shed traces of itself downstream. The team ran proteomics, lipidomics and metabolomics across urine from patients and healthy controls, and found the disease leaves a consistent metabolic fingerprint. Out of that they distilled something practical: three proteins, SAA1, haptoglobin and LCN15, combined into a score they call the UrineScore that flags early-stage disease off a fast, targeted mass-spectrometry assay.

The appeal is in what it removes. Today a suspicious kidney is confirmed by imaging and, often, biopsy, an invasive step with its own risks and delays. A urine test asks nothing of the patient beyond what they already produce, exactly what a screening tool needs to reach the people now diagnosed too late, or by luck. It would also give oncologists a low-cost way to watch known patients for recurrence without repeat scans.

A biomarker is not yet a screening programme. Three proteins that separate patients from controls in a study cohort still have to hold up across the far messier population of everyone, where other conditions muddy the same signals, and that validation is the work ahead. But the logic is sound and the sample is free, a rare combination in early cancer detection. The kidney, it turns out, was reporting on itself the whole time. Someone had to collect the message.

→ Jonsson, Oliveira, Hofmann et al., EMBO Molecular Medicine, 2026

Radar Scans

Why colon cancer shrugs off immunotherapy, and how to change that

Immune checkpoint therapy has transformed several cancers but does almost nothing for most metastatic colorectal cancer, and the reason has been murky. A study in Cell Death and Differentiation, led by Veronica Moreno-Viedma and Emi Adachi Fernandez in Maria Sibilia 's group at the Medical University of Vienna with colleagues at IRB | Institute for Research in Biomedicine Barcelona, points at the immune cells around the tumour rather than the tumour itself. The receptor EGFR, long known as a driver on cancer cells and already a drug target, turns out to sit on granulocytes in the tumour's surroundings too, where it helps hold the immune response down. Switch it off in those cells, and in a mouse model the body's defences re-engage and checkpoint therapy starts to work. That is preclinical evidence, not the clinic, but it recasts a familiar target: the same receptor does one job on the cancer cell and another on the immune cell beside it.

→ Moreno-Viedma, Adachi-Fernandez, Krauß et al., Cell Death and Differentiation, 2026

What TikTok is worth to a song depends on the song

Does TikTok drive people to paid streaming, or eat into it? Earlier studies split both ways, and a Vienna led team shows why: they were measuring different things. Daniel Winkler and Christian Hotz-Behofsits at WU (Vienna University of Economics and Business), with Nils Wlömert and international co-authors, use Universal Music Group's 2024 pull of its whole catalogue from TikTok as a natural experiment. Music demand is wildly top-heavy, the top tenth of songs makes 96% of TikTok clips and 76% of Spotify streams, so a method that quietly averages across the long tail hides the effect that matters. The study finds TikTok's promotional pull is real but almost all of it sits in the viral hits. For everything below, losing TikTok barely registers.

→ Winkler, Hotz-Behofsits, Wlömert et al., Quantitative Marketing and Economics, 2026

Can a crowd owe you something?

Strangers on a beach could together save a drowning child that none of them could reach alone. Does the group then have a duty to act, even though a crowd is not a person, has no mind, and decides nothing? Niels de Haan, a philosopher at the University of Vienna, takes up this Puzzle of Collective Duties and defends a principle others drop to escape it: that only a moral agent can bear a moral duty. Abandon it, he argues, and no principled line survives between a crowd and a dog or a houseplant as a duty-bearer. It is not an idle worry. Whether humanity as a whole can be said to owe action on poverty or climate is exactly this question, and de Haan's answer is that assigning the duty to an entity that cannot grasp it may say less than it seems to.

→ de Haan, Synthese, 2026

The genetic switch that sets a jellyfish loose

An Aurelia jellyfish starts as a polyp fixed to the seabed, then rebuilds itself into the free-swimming medusa everyone knows. What changes at the level of cells during that switch had never been mapped. A University of Vienna team around Alison Cole and Ulrich Technau, with Kristin Janicek among the lead authors, tracked it cell by cell in PLoS Biology and found the medusa carries a richer cell inventory than the polyp: more neuron types, and striated muscle, the tissue a jellyfish needs to swim, appearing where the polyp had none. Two of those neuronal lineages are built by the same transcription factors as in the sea anemone Nematostella, a relative that split off some 500 million years ago. That points to gene duplication as one old engine behind the growing cellular complexity of this branch of animals.

→ Link, Jahnel, Janicek et al., PLoS Biology, 2026

One legendary flood turns out to have been sixteen

The St. Mary Magdalene's Flood of July 1342 has stood in European memory as the worst deluge of the last millennium, one catastrophe that drowned much of Central Europe. A Nature study led by Andrea Kiss and Gunter Bloschl at the TU Wien reads that memory against more than a thousand medieval chronicle entries, tree rings and sediment records, and finds it was the peak of a chain of sixteen major floods between late 1341 and 1343. Four of them were once-in-500-to-1,000-year events on their own. Modern flood defences are built largely for a single rare event, not for catastrophes that arrive back to back before a region has recovered, which is the harder case the record now hands us.

→ Kiss, Viglione, Barriendos et al., Nature, 2026

WWTF Insight

Two WWTF-funded papers turned up in this fortnight's Crossref data, both from Medical University of Vienna, both about seeing something standard practice misses. Rainer Oberbauer's nephrology group reports that matching a kidney donor and recipient on HLA tissue type, the long-standing basis for allocation, captures only part of what decides whether a graft lasts. Screening the rest of the genome for donor-recipient mismatch adds real predictive power, and the group folds both into one compatibility score meant to spare low-risk patients from over-immunosuppression and flag high-risk ones earlier (LS20-081). Saiedeh Saghafi, separately, shows that adding ultrasound to the chemical clearing that turns a whole organ transparent for 3D imaging speeds it up enough to handle specimens once too dense to process (NXT22-001). One reads a genome for a risk HLA typing leaves invisible; the other makes the inside of an intact organ visible at all. WWTF backed both.

Stray Signal

Mahler's Seventh, filed as data

The first movement of Mahler's Seventh has a reputation for being hard to hold together, and conductors have pulled its tempo and colour in wildly different directions for a century. Christian Utz, a musicologist at the University of Vienna, studies exactly that spread across dozens of recordings, part statistics, part close listening. What is unusual is what he did with the evidence: he put the sound itself on Zenodo, nine audio examples from historic recordings plus a few Sonic Visualiser renderings, as a citable, openly licensed dataset. Analysis of performance usually leaves only prose behind, the recordings locked away and the listening lost once the article is done. Here the thing under analysis stays on the record for anyone to check.

→ Utz, Zenodo, 2026