Mirror Molecule Kills Cancer, Spares Healthy Cells

A Molecular Mirror Image Takes Aim at Tumors

In a finding that could reshape cancer treatment, an international team led by researchers at the University of Geneva (UNIGE) and the University of Marburg has demonstrated that D-cysteine — a rare, mirror-image version of the common amino acid cysteine — can dramatically slow tumor growth while leaving healthy cells unharmed. The study was published in Nature Metabolism in August 2025.

Cysteine exists in two molecular forms that are near-identical but mirror images of each other, like left and right hands. The "L" form is used by living organisms; the "D" form is rare in biology. That difference, it turns out, is precisely what makes D-cysteine so promising as an anticancer agent.

How It Works: Exploiting a Cancer Cell's Weakness

Many tumors overexpress a surface protein called xCT/CD98, a transporter that cancer cells rely on to import cystine and fuel their antioxidant defenses. This same transporter mistakenly absorbs D-cysteine — smuggling the mirror molecule inside the tumor cell.

Once inside, D-cysteine blocks NFS1, a mitochondrial enzyme essential for building iron-sulfur clusters. These tiny structures are indispensable for cellular respiration, DNA replication, and genome stability. By shutting NFS1 down, D-cysteine triggers DNA damage, stalls the cell cycle, and halts tumor growth.

Crucially, healthy cells that do not overexpress xCT/CD98 are largely unable to import the molecule and remain unaffected. In mouse models of aggressive mammary cancer, tumor growth slowed markedly with no major side effects observed — a striking contrast to conventional chemotherapy, which routinely damages healthy tissue.

"The selectivity is the key innovation here," the Geneva researchers noted. "We are exploiting a vulnerability that is specific to the cancer cell's own survival strategy."

AI Reads Brain MRIs in Seconds

In a parallel leap forward, engineers at the University of Michigan have unveiled Prima, an AI foundation model capable of interpreting brain MRI scans in seconds. The results, published in Nature Biomedical Engineering in February 2026, show that Prima achieved a mean diagnostic accuracy of 92.0% AUC across 52 neurological conditions — reaching up to 97.5% accuracy on specific diagnoses.

Prima is a vision-language model trained on more than 220,000 MRI studies encompassing 5.6 million imaging sequences, combined with patients' clinical histories and physician-indicated reasons for each scan. In a year-long validation study covering nearly 30,000 MRI cases at Michigan Health, Prima outperformed all other state-of-the-art AI models tested.

Beyond diagnosis, Prima can triage urgency and recommend the appropriate specialist — whether a stroke neurologist or a neurosurgeon — immediately after a patient completes imaging. In settings where radiology backlogs can delay life-saving interventions, this capability alone could prove transformative.

Two Breakthroughs, One Direction

Though distinct in their science, both discoveries point toward the same horizon: medicine that is more precise, more targeted, and less harmful. D-cysteine represents a new class of metabolic attack on cancer — one engineered around a tumor's own molecular greed. Prima demonstrates that AI can now serve as a reliable frontline diagnostic partner for neurologists overwhelmed by imaging volume.

Neither technology is yet in clinical use. D-cysteine must advance through human trials before reaching patients, and Prima requires further validation with richer patient data. But both signal that the era of blunt, one-size-fits-all medicine is giving way to something far more sophisticated — and far more humane.