Mines Paris, PSL University

Immunotherapy is improving the survival of patients with metastatic non-small cell lung cancer (NSCLC), yet reliable biomarkers are needed to identify responders prospectively and optimize patient care. In this study, we explore the benefits of multimodal approaches to predict immunotherapy outcome using multiple machine learning algorithms and integration strategies. We analyze baseline multimodal data from a cohort of 317 metastatic NSCLC patients treated with first-line immunotherapy, including positron emission tomography images, digitized pathological slides, bulk transcriptomic profiles, and clinical information. Testing multiple integration strategies, most of them yield multimodal models surpassing both the best unimodal models and established univariate biomarkers, such as PD-L1 expression. Additionally, several multimodal combinations demonstrate improved patient risk stratification compared to models built with routine clinical features only. Our study thus provides evidence of the superiority of multimodal over unimodal approaches, advocating for the collection of large multimodal NSCLC datasets to develop and validate robust and powerful immunotherapy biomarkers.

Suppressing deep-level defects at the perovskite bulk and surface is indispensable for reducing the non-radiative recombination losses and improving efficiency and stability of perovskite solar cells (PSCs). In this study, two Lewis bases based on chalcogen-thiophene (n-Bu4S) and selenophene (n-Bu4Se) having tetra-pyridine as bridge are developed to passivate defects in perovskite film. The uncoordinated Pb²⁺ and iodine vacancy defects can interact with chalcogen-concave group and pyridine group through the formation of the Lewis acid-base adduct, particularly both the defects can be surrounded by concave molecules, resulting in effective suppression charge recombination. This approach enables a power conversion efficiency (PCE) as high as 25.37% (25.18% certified) for n-i-p PSCs with stable operation at 65 °C and 1-sun illumination for 1300 hours in N2 (ISOS-L-2 protocol), retaining 94% of the initial efficiency. Our work provides insight into the bowl-shaped Lewis base in defects passivation by coordinated strategy for high-performance photovoltaic devices.

Accurate very short‐term solar irradiance forecasting is crucial for optimizing the integration of solar energy into power systems. Herein, an image‐based deep learning framework for minute‐scale solar irradiance prediction is presented. The locally developed model is benchmarked against two commercial forecasting solutions deployed at the same experimental site, demonstrating superior accuracy and adaptability. A key contribution is the introduction of a skill‐driven sampling algorithm based on clear sky index persistence error, which optimizes the training dataset by excluding low‐utility samples while retaining essential physical features like solar zenith and azimuth angles. This algorithm enables the exclusion of up to 30% of the original training data, resulting in ≈16% savings in computational resources without affecting forecast accuracy validated using a test set of 324 991 observations. The model achieves a skill score of 7.63%, significantly outperforming the commercial models, which exhibit negative skill scores under the same conditions.

Transcription factors are frequent cancer driver genes, exhibiting noted specificity based on the precise cell of origin. We demonstrate that ZIC1 exhibits loss-of-function (LOF) somatic events in group 4 (G4) medulloblastoma through recurrent point mutations, subchromosomal deletions and mono-allelic epigenetic repression (60% of G4 medulloblastoma). In contrast, highly similar SHH medulloblastoma exhibits distinct and diametrically opposed gain-of-function mutations and copy number gains (20% of SHH medulloblastoma). Overexpression of ZIC1 suppresses the growth of group 3 medulloblastoma models, whereas it promotes the proliferation of SHH medulloblastoma precursor cells. SHH medulloblastoma ZIC1 mutants show increased activity versus wild-type ZIC1, whereas G4 medulloblastoma ZIC1 mutants exhibit LOF phenotypes. Distinct ZIC1 mutations affect cells of the rhombic lip in diametrically opposed ways, suggesting that ZIC1 is a critical developmental transcriptional regulator in both the normal and transformed rhombic lip and identifying ZIC1 as an exquisitely context-dependent driver gene in medulloblastoma.

We investigate the spreading of falling ambient-temperature Newtonian drops after their normal impact on a quartz plate covered with a thin layer of liquid nitrogen. As a drop expands, liquid nitrogen evaporates, generating a vapour film that maintains the drop in levitation. Consequently, the latter spreads in inverse Leidenfrost conditions. Three drop-spreading regimes are observed: (i) inertio-capillary, (ii) inertio-viscous, and (iii) inertio-viscous-capillary. In the first regime, although the drop expansion is essentially driven by a competition between inertial and capillary stresses, it is also affected by viscous effects emerging from the vapour film, which ultimately favours the development of a shear flow within the drop. Interestingly, vapour film effects become marginal in both the second and third regimes, allowing the drop to undergo biaxial extension primarily. More specifically, in the inertio-viscous scenario, the expansion is driven by the balance between inertial and biaxial extensional viscous stresses in the drop. Finally, inertia, capillarity and drop viscosity are all relevant in the third regime. These physical mechanisms are underlined through a mixed approach combining experiments with multiphase three-dimensional numerical simulations in light of spreading dynamics analyses, energy transfer and scaling laws. Our results are rationalized in a two-dimensional diagram linking the drops’ maximum expansion and spreading time with the observed spreading regimes through a single dimensionless parameter given by the square root of the capillary number (the ratio of the viscous stress to the capillary stress).

We develop and implement a new method for identifying wasted subsidies and use it to provide systematic evidence of the misallocation of carbon offsets in the Clean Development Mechanism—the world's largest carbon offset program. Using newly constructed data on the locations and characteristics of over 1,000 wind farms in India, we estimate that at least 52 percent of approved carbon offsets were allocated to projects that would very likely have been built anyway. We estimate that the sale of these offsets to regulated polluters resulted in substantially higher global carbon dioxide emissions.(JEL H23, O13, Q42, Q54, Q58)

The operation of large-scale infrastructure networks requires scalable optimization schemes. To guarantee safe system operation, a high degree of feasibility in a small number of iterations is important. Decomposition schemes can help to achieve scalability. In terms of feasibility, however, classical approaches such as the alternating direction method of multipliers (ADMM) often converge slowly. In this work, we present primal decomposition schemes for hierarchically structured strongly convex QPs. These schemes offer high degrees of feasibility in a small number of iterations in combination with global convergence guarantees. We benchmark their performance against the centralized off-the-shelf interior-point solver Ipopt and ADMM on problems with up to 300,000 decision variables and constraints. We find that the proposed approaches solve problems as fast as Ipopt, but with reduced communication and without requiring a full model exchange. Moreover, the proposed schemes achieve a higher accuracy than ADMM.

Efficient drilling operations require optimal drilling parameters to achieve higher penetration rates and minimize tool wear. This study focuses on characterizing the piston coefficient of restitution (COR) as a damage indicator for rocks dynamically loaded by percussive tools. The COR offers a nuanced understanding of damage, particularly at low impact energies where plastic deformation beneath the insert and sub-surface fractures predominates over chip mass removal. Various design parameters, including piston-to-bit mass ratio, piston length and impact velocity, were varied to evaluate their impact on the piston COR in pristine rock samples. Tests were conducted on granite, sandstone and limestone, all common rocks in down-the-hole hammer drilling. Despite variations in the impact energy level, the influence of mass or length ratios on the COR and the resultant destroyed rock volume was minimal. X-ray computed tomography revealed the significance of the crushed zone beneath the crater, affecting the rock ability to exert mechanical work on the bit during unloading. This characteristic led to the identification of a distinctive non-linear shape in the COR curves, with clearly delineated regions for each rock type, corresponding to different damage rates. These findings underscore the potential of the piston COR as a valuable tool for aiding in the identification of optimal drilling parameters and understanding rock formation characteristics during drilling.

The Ségognole 3 shelter lies within a quartzitic sandstone megaclast in a lag deposit in the Paris Basin. It displays a female sexual configuration associated with a horse engraving, stylistically attributed to the Upper Palaeolithic. Recent studies have demonstrated that modifications to the natural features of the shelter had been undertaken to cause water to flow through what is seen as the vulva. New investigations reported here describe additional modifications to natural features in the shelter to direct rainwater infiltration to a network of channels engraved onto the shelter floor to form a functioning representation of watercourses. The carved motifs and their relationship with natural features in the sandstone of the shelter can be compared with major geomorphological features in the surrounding landscape. The engraved floor is not quite a map but more like a model in miniature of the surrounding landscape, potentially a world‐first 3D‐model of a Palaeolithic territory.

An efficient one‐step extraction method was developed for the recovery of additives and non‐intentionally added substances (NIAS) from polystyrene, performed at room temperature for 2.33 h, without grinding to avoid fostering the formation of NIAS unrelated to polymer processing. Solvent use (39.2 mL/g) was greatly reduced compared with extraction conditions previously reported. The study of NIAS is analytically challenging but with high importance since their presence is a potential threat to human health and to the environment while reducing plastic potential recyclability. For an understanding of NIAS formation and influence of processing parameters, a systematic approach was taken, using virgin polystyrene mixed with known quantities of standard additives as model materials (Irganox_1076, Tinuvin_326, Irgafos_168). The degradation of one additive was identified by NMR and GC–MS. Precise multiple‐point quantification with internal standard was performed by GC–MS, measuring a 5.1 wt% Irgafos_168 degradation, with additives LOD ranging from 0.55–0.95 ppm. Evaluation of analytical challenges, such as matrix effects, was discussed and quantified. This method will help the quality control of virgin and recycled PS materials, including food contact ones, and improve the knowledge of PS processing impact on NIAS formation.

Remaining Useful Life (RUL) predictions is a key technology for device prognostic and health management. Due to deficiencies in data and models during the prediction process, the predicted RUL results exhibit various types of uncertainty. However, most RUL prediction models address point estimates or total uncertainty. To this end, this paper proposes a Bayesian data-driven RUL framework with aleatoric uncertainty and epistemic uncertainty quantification. First, considering the impact of data inherent noise and model ignorance on prediction uncertainty separately, an algorithm for quantifying aleatoric and epistemic uncertainty of Relevance Vector Machine is proposed by Monte Carlo sampling. Then a Bayesian data-driven RUL predictive framework with uncertainty quantification is proposed. Adaptive training set based on the similarity method is adopted to extract units of training set with features are similar to the test unit. Finally, the application of the proposed framework is shown on a public turbofan engine dataset C-MAPSS and a case of the Once-Through Steam Generator of nuclear power plants. The superior prediction performance of the proposed framework is illustrated by comparing with other state-of-art methods.

In this work, we demonstrate that initially misoriented gallium nitride (GaN) crystalline grains grown on top of GaN/AlN/Si/SiO2 nano-pillars, and which have nucleated independently, realign themselves upon coalescence to form high crystalline quality GaN platelets. Electron backscatter diffraction (EBSD) combined with cathodoluminescence (CL) and scanning x-ray diffraction microscopy (SXDM) provided complementary information on the structural properties of GaN before and during the initial coalescence growth phase. SXDM measurements on GaN coalescing at an early growth stage and on GaN pillars only (prior to growth) confirmed that the initially misoriented GaN pillars coalesce into larger well-defined GaN domains (3.9 μm) very well oriented by themselves, with a spatially varying broadening of the diffraction peak that is maximum at the boundaries between neighboring domains, as identified in the spatially resolved orientation maps. The presence of geometrically necessary dislocations (GNDs) at the domain boundaries detected in the EBSD is confirmed by CL images and the estimated GND density is 2 × 1011 cm⁻² in these specific zones. Additionally, statistical analysis of SXDM maps indicated that 0.1° of tilt between neighboring pillars constitutes the limit for the current pendeo-epitaxy growth approach for the formation of pillar groups similar in size to the perfectly aligned GaN domains upon coalescence. This work illustrates the potential of this growth strategy to produce high crystalline quality GaN platelets adapted for micro-LEDs growth, and, most importantly, it provides a microscopic insight into the coalescence process, which could be extended to other materials and growth approaches.

Illuminated medieval manuscripts are of outstanding value and their preservation is of great importance, not only because of their beauty but also because of the information they contain about medieval society. This work focuses on the evaluation of the parchment's state of preservation of the Prayer book of Mary of Guelders, which comprises about 600 folios. The knowledge gained should support the decision-making process regarding suitable conservation measures. An assessment of the preservation state of the parchment was performed from the macro- down to the microscale. Optical observations of cracks in the parchment and colour measurements preceded chemical analyses. The hydrothermal stability of the fibres was evaluated by means of observations using a micro hot table (MHT). The chemical state of preservation of parchment was evaluated using Laboratory-based Fourier transform (FT) Infrared (IR) analysis in reflection mode as well as synchrotron FTIR imaging in transmission mode at the IRIS beamline at BESSY II/ HZB in Berlin. The study allowed the conclusion that the parchment of the Prayer book of Mary of Guelders was in good state of preservation and indicated that the parchment changes were mainly caused by mechanical stress on the folios due to tight binding of the book and not by chemical processes.

Aluminum alloys are light and corrosion-resistant materials, which is why they are widely used in structures in many industrial fields (construction, automotive, electric cables). The article deals with the aluminum busduct structure. Therefore, the mechanical and especially electrical properties of busduct welds are the basic criteria for assessing the quality of welds. The aim of the work was to present the advantages of a process combining metal inert gas welding with immediate microjet cooling (MJC). The parameters of aluminum welding using the micro-jet method were estimated in order to obtain products with the desired strength, mechanical and electrical parameters. Information regarding the influence of various microjet parameters on the metallographic structure was also recorded. Then, the metallographic properties and some physical properties of the welding structures (mechanical resistance, electrical conductivity) were examined. In addition, computer simulations of the welding process with micro-jet cooling were performed. The heat affected zone in the welded material was determined. The proposed numerical method will allow the assessment of the parameters of the welding process with micro-jet cooling depending on the parameters of the materials undergoing the welding process. The numerical approach will significantly reduce costly and time-consuming in situ work. Planning the welding of large structures (such as busducts) will be more economical using the results of computer simulations.