The massive swarm of earthquakes that rattled the Greek islands of Santorini and Amorgos in 2025 was not caused by a slipping fault – it was triggered by pulses of magma tunneling far below the seafloor, according to a new study. The findings offer a detailed look at a “pumping” magmatic dike in action and provide a foundation for more reliable, physics-based eruption forecasting and volcanic hazard assessment. In early 2025, a burst of intense earthquakes – including several around magnitude 5 – shook the region between the islands of Santorini and Amorgos in the Aegean Sea. Because Santorini is an active volcano with a history of catastrophic eruptions, the event raised serious concerns. Exactly what triggered this seismic unrest remains debated, but it is generally attributed to magmatic dike intrusion or fluid-driven tectonic fault slip. However, fully determining the processes that contributed to the event is difficult to resolve because most magmatic dike activity occurs deep underground or far offshore, beyond the scope of traditional monitoring methods.

To overcome these limits, Anthony Lomax and colleagues applied machine learning methods to detect and precisely locate ~25,000 earthquakes recorded during the 2025 Santorini-Amorgos. By applying a new three-dimensional imaging technique, CoulSeS, which treats earthquake locations and indicators of stress change at depth, Lomaxz et al. were able to use the tremblors as “virtual sensors” to map the underlying geologic source of the unrest. By modeling how evolving patterns of stress triggered seismic activity and tracing how earthquakes migrated, the authors found that the event was driven by the intrusion of a horizontally propagating magma-filled dike, which extended about 30 kilometers below the seafloor between the two islands. High-resolution imaging revealed a complex pattern of pressure fluctuations – as the dike propagated, it repeatedly broke through stress barriers in the crust, surged forward, and then underwent cycles of contraction and expansion, creating a dynamic pumping behavior that earlier studies had overlooked. “The study of Lomax et al. could lead to new dynamic models of magma transport that account for spatial variations in the fracture resistance of surrounding rocks,” writes Virginie Pinel in a related Perspective. “Furthermore, combining real-time observations and dynamic models could predict the location and timing of eruptions by using advanced data assimilation or machine-learning techniques."

This study presents a comprehensive benchmarking of human-derived transcription activation domains (hTADs) for use in CRISPR-based artificial transcription factors (ATFs), aiming to circumvent the immunogenicity associated with viral TADs like VP64 and VPR. The authors systematically compared eight hTADs—CITED1, CITED2, MYB, KLF7, CSRNP1, NFZ, MSN, and p65HSF1—fused to dCas9. While several hTADs surpassed VP64 in activating a synthetic EGFP reporter, none outperformed VPR at endogenous gene targets.

A new study shows that an AI assistant can conduct assessment conversations with patients with higher accuracy than the rating scales used in healthcare today. In the study, 303 participants were interviewed by the AI ​​assistant Alba, who then suggested possible psychiatric diagnoses.