TUB-010 is a next-generation antibody-drug conjugate (ADC) targeting CD30 expressed on various hematopoietic malignancies such as Hodgkin lymphoma. Among the therapeutic options for patients with relapsed and refractory CD30-positive cancers is brentuximab vedotin (Adcetris), a monomethyl auristatin E (MMAE)-delivering anti-CD30 ADC with a mean drug-to-antibody ratio of 4. Adcetris exhibits a high response rate at the cost of significant toxicities, likely driven by the payload MMAE and the instability of the maleimide conjugation chemistry. TUB-010 uses the same antibody and payload as Adcetris but is based on the Tub-tag conjugation strategy, which stably attaches MMAE to the hydrophilic Tub-tag peptides on the light chains via chemoenzymatic conjugation. This new technology enables the generation of a homogeneous and site-specific drug-to-antibody ratio 2 ADC with unique biophysical properties. TUB-010 demonstrates similar binding and lysosomal release characteristics as Adcetris, which translates into comparable in vitro cytotoxicity on CD30-positive cell lines when normalized to the MMAE concentration. Importantly, TUB-010 exhibits higher stability with negligible premature deconjugation in circulation and reduced aggregation, as well as lower nonspecific cytotoxicity on target-negative cells compared with Adcetris. As a consequence, TUB-010 induces superior tumor control compared with Adcetris when dosed at equal MMAE concentrations in vivo and also lower toxicity and higher tolerability in rodents and nonhuman primates. Taken together, TUB-010 is a novel, potential best-in-class anti-CD30 ADC with improved biophysical properties designed to deliver MMAE with higher precision and a wider therapeutic window than Adcetris using Tub-tag technology. Therefore, TUB-010 may increase the clinical benefit of anti-CD30 ADC therapies.
Wildfires, insect outbreaks, and storms cause large pulses of tree mortality. Climate change amplifies these forest disturbances, yet their future magnitude and extent remain uncertain. Here, we simulated future forest disturbance regimes at 100-meter resolution across Europe using a deep learning-based simulation framework. Our results show that forest disturbances will continue to increase throughout the 21st century, with disturbed areas more than doubling relative to the recent past under an unabated continuation of climate change. Wildfires are the main agent driving future disturbance change. Changing disturbances result in an increase in young forests, substantially altering Europe's forest demography. Because of their profound implications for forest carbon storage and the habitat value of forest ecosystems, disturbances should be a priority of forest policy and management.
Quantifying animal movements is necessary for answering a wide array of research questions in ecology and conservation biology. Consequently, ecologists have made considerable efforts to identify the best way to estimate an animal’s home range, and many methods of estimating home ranges have arisen over the past half century. Most of these methods fall into two distinct categories of estimators that have only recently been described in statistical detail: those that measure range distributions (methods such as Kernel Density Estimation that quantify the long-run behavior of a movement process that features restricted space use) and those that measure occurrence distributions (methods such as Brownian Bridge Movement Models and the Correlated Random Walk Library that quantify uncertainty in an animal movement path during a specific period of observation). In this paper, we use theory, simulations, and empirical analysis to demonstrate the importance of applying these two classes of space use estimators appropriately and distinctly. Conflating range and occurrence distributions can have serious consequences for ecological inference and conservation practice. For example, in most situations, home-range estimates quantified using occurrence estimators are too small, and this problem is exacerbated by ongoing improvements in tracking technology that enable more frequent and more accurate data on animal movements. We encourage researchers to use range estimators to estimate the area of home ranges and occurrence estimators to answer other questions in movement ecology, such as when and where an animal crosses a linear feature, visits a location of interest, or interacts with other animals. Open Research Statement Tracking data on Aepyceros melampus, Beatragus hunteri, Bycanistes bucinator, Cerdocyon thous, Eulemur rufifrons, Glyptemys insculpta, Gyps coprotheres, Madoqua guentheri, Ovis canadensis, Propithecus verreauxi, Sus scrofa, and Ursus arctos are publicly archived in the Dryad repository (Noonan et al. 2018; https://doi.org/10.5061/dryad.v5051j2), as are data from Procapra gutturosa (Fleming et al. 2014a; https://doi.org/10.5061/dryad.45157). Data on Panthera onca were taken from (Morato et al. 2018). Additional data are publicly archived in the Movebank repository under the following identifiers: Canis latrans, 8159699; Canis lupus, 8159399; Chrysocyon brachyurus, 18156143; Felis silvestris, 40386102; Gyps africanus, 2919708; Lepus europaeus, 25727477; Martes pennanti, 2964494; Panthera leo, 220229; Papio cynocephalus, 222027; Syncerus caffer, 1764627; Tapirus terrestris, 443607536; Torgos tracheliotus, 2919708; and Ursus americanus, 8170674.
Landslides present a critical hazard in the Himalayas, where steep topography, intense rainfall, and tectonic activity converge to destabilize slopes. Accurate delineation of high-susceptibility zones is essential to safeguard lives, infrastructure, and ecosystems. Here, we construct a comprehensive Landslide Susceptibility Map (LSM) for Uttarakhand, a landslide-prone state in northern India, by integrating advanced ensemble machine learning (ML) with explainable AI. Our analysis comprises 35 geo-environmental variables, ranging from historical landslide inventories and remote sensing data to GIS-based geomorphological, hydrological, and anthropogenic layers. We evaluate six ML models (Logistic Regression, Support Vector Machine, Random Forest, Extra Trees, Gradient Boosting, and eXtreme Gradient Boosting) before consolidating them into a stacking ensemble (SE), achieving an Area Under the Curve (AUC) of 0.987 on the training set and 0.979 on the test set. Across models, false-negative rates were low; Extra Trees minimized missed events (FNR = 3.5 %) but with a high false-positive rate (23.6 %), whereas XGBoost and the SE achieved a better sensitivity-specificity balance (FNR = 5.6 and 5.5 %, respectively) with comparatively lower false positives, favoring operational use. Spatial transferability to Sikkim was strong (Uttarakhand test accuracies 0.864-0.917; Sikkim 0.905-0.971), with XGBoost yielding the highest Sikkim test accuracy (0.971) and ensemble approaches (GB, XGBoost, SE) all exceeding 0.96, highlighting robust generalization across different Himalayan regions. Our ensemble model surpasses all individual models and classifies the study area into five susceptibility zones (very low to very high), with 18.20 % of Uttarakhand, particularly in Pithoragarh, Chamoli, and Rudraprayag districts, falling under high-susceptibility zones. Further interpretability is provided by SHapley Additive exPlanations (SHAP), which highlight key drivers of slope failure, including slope angle, fault proximity, and rainfall. Our findings highlight the value of combining robust ML techniques with geoscientific data, thereby enhancing hazard assessments and informing disaster risk reduction across the Himalayas and similarly vulnerable terrains worldwide.
Shifting consumers towards more sustainable food consumption and avoiding food waste have been identified as key levers in mitigating food systems-related climate change impacts. Here we conducted a machine-learning-assisted systematic review and meta-analysis of 306 effect sizes from 110 articles, covering over 2.4 million observations, to assess the effectiveness of demand-side interventions targeting actual or incentivized behaviours. On average, we find small effect sizes across both food consumption and food waste interventions. Effect sizes vary substantially across intervention types, with certain choice architecture interventions, such as availability and defaults, driving much of the overall effect in both domains, while incentives also show promise in reducing food waste. These effects remain robust even after accounting for severe publication bias, which notably reduces average estimates for other intervention types. Sensitivity analyses further underscore the need for future research to systematically identify when, how and why interventions are effective. Demand-side food policies show small average effects, with effectiveness varying across intervention types. Targeted strategies and a better understanding of how context and population influence outcomes can advance food system sustainability.
