Partial Dependence Research Articles

Book-ahead ride-hailing trip and its determinants: Findings from large-scale trip records in China

The cruising of ride-hailing vehicles generates negative externalities such as traffic congestion and vehicular emissions. These externalities can be mitigated by reducing cruising driving via operating book-ahead ride-hailing services, where the platform dispatches and routes drivers based on precise information on travelers’ departure time and origin–destination (OD). However, the effects of factors influencing book-ahead ride-hailing trips have rarely been empirically examined with real data. Using six-month trip data from China, this study employs a gradient boosting decision tree (GBDT) method with hyperparameters optimized by the Bayesian optimization algorithm to examine the factors associated with book-ahead ride-hailing trips across OD pairs (hexagon cells-to-hexagon cells) at various spatial scales. The relative importance rankings generated from this study indicate that trip features, weather conditions, and accessibility to transportation hubs are significant determinants correlated with the usages of book-ahead ride-hailing. The partial dependence plots demonstrate the nonlinear threshold effects of these determinants on the hourly number of book-ahead ride-hailing per OD pair. Moreover, this study compares the differences in associations between peak and non-peak hours as well as weekdays and weekends. The disparity in the nonlinear threshold effects between weekdays and weekends is only observable during the evening peak period and not at other times. These findings provide valuable insights into developing practical strategies for promoting book-ahead ride-hailing services.

Ischemia‒Reperfusion accelerates neointimal hyperplasia via IL-1β-mediated pyroptosis after balloon injury in the rat carotid artery

BackgroundIschemia‒reperfusion (IR) is a pathological process that causes secondary damage to blood vessels. However, whether IR can further worsen neointima formation after balloon injury and the detailed mechanism are unclear. MethodsAn in vivo model of balloon injury to the rat carotid artery was established to study the effect of IR following balloon injury on neointima formation. Smooth muscle cells (SMCs) were isolated from rat aortas and exposed to hypoxia-reoxygenation to mimic the IR process in vitro. The in vitro cell model was used to investigate the mechanism of IR-mediated neointima formation after balloon injury, which was further confirmed in an in vivo rat model. ResultsIR aggravated neointima formation in the rat carotid artery 2 weeks after balloon injury compared with that observed in the absence of balloon injury (P < 0.001). Compared with that of normal SMCs in the rat carotid artery, the expression of IL-1β, a key proinflammatory cytokine associated with pyroptosis, was increased more than 3-fold in the IR-induced neointima (P < 0.0001) and contributed to the proliferation and migration of rat primary aortic SMCs (P < 0.0001). This process was alleviated by the antioxidant acetylcysteine (NAC), suggesting its partial dependence on intracellular ROS. In the rat model of IR following balloon injury in the carotid artery, the carotid artery that was locally transfected with AAV carrying sh-IL-1β or sh-caspase-1, which alleviated neointima formation, as indicated by a reduction in intima-media thickness in the rat carotid artery (P < 0.0001). ConclusionOur results suggested that IR could promote IL-1β production in SMCs in the carotid artery after balloon injury and aggravate neointimal hyperplasia, which was alleviated by silencing caspase-1/IL-1β signaling in SMCs in the carotid artery. These results suggest that IL-1β may be an effective target to combat IR-related neointima formation.

Examining Commercial Crime Call Determinants in Alley Commercial Districts before and after COVID-19: A Machine Learning-Based SHAP Approach

Although several previous studies have examined factors influencing crime at a specific point in time, limited research has assessed how factors influencing crime change in response to social disasters such as COVID-19. This study examines factors, along with their relative importance and trends over time, and their influence on 112 commercial crime reports (illegal street vendors, dining and dashing, minor quarrels, theft, drunkenness, assault, vagrancy and disturbing the peace) in Seoul’s alley commercial districts between 2019 and 2021. Variables that may affect the number of commercial crime reports are classified into four characteristics (socioeconomic, neighborhood, park/greenery and commercial district attributes), explored using machine learning regression-based modeling and analyzed through the use of Shapley Additive exPlanations to determine the importance of each factor on crime reports. The Partial Dependence Plot is used to understand linear/non-linear relationships between key independent variables and crime reports. Among several machine learning models, the Extra Trees Regressor, which has the highest performance, is selected for the analysis. The results show a mixture of linear and non-linear relationships with the increasing crime rates, finding that store density, dawn sales ratio, the number of gathering facilities, perceived urban decline score, green view index and land appraisal value may play a crucial role in the number of commercial crimes reported, regardless of social trends. The findings of this study may be used as a basis for building a safe commercial district that can respond resiliently to social disasters.

Study on waste tire pyrolysis product characteristics based on machine learning

Tire pyrolysis is a highly complex thermochemical conversion process that transforms waste tires into high-value products such as pyrolysis oil, pyrolysis gas, and pyrolysis char. This process significantly mitigates the environmental issues caused by waste tires and reduces reliance on fossil resources. The physicochemical properties of tires and pyrolysis operation parameters have a significant impact on the yield of the three-phase products, thus affecting the industrial viability of tire pyrolysis to a large extent. Traditional prediction methods such as computational fluid dynamics and process simulation often fail to provide satisfactory results. However, data-driven machine learning (ML) models have demonstrated their ability to handle complex nonlinear problems and offer more reliable predictions of pyrolysis products yield. This study employed a collected database of tire pyrolysis to develop tire pyrolysis product prediction models based on five ML models. These models were further optimized using Particle Swarm Optimization (PSO), and their prediction performances were quantitatively evaluated to identify the optimal model. Shapley analysis and one-way partial dependence analysis were conducted to explore the impact of input features on the output responses. Furthermore, an industrial-grade software was developed for accurate prediction of tire pyrolysis three-phase products yield. The results revealed that Gaussian process regression (GPR) and random forest regression (RFR), both optimized with PSO, demonstrated impressive prediction performance. Among them, the GPR model achieved the highest prediction accuracy with coefficient of determination (R2) values of 0.964, 0.924, and 0.86 for oil, char, and gas yields respectively, during the testing stage.

Examining temporally varying nonlinear effects of urban form on urban heat island using explainable machine learning: A case of Seoul

Many empirical studies have examined the relationship between urban form and canopy layer urban heat island (CUHI). However, these studies mostly have used small sample sizes and statistical methods that assume CUHI responds linearly to urban form features. Additionally, the differences in the effect of urban form on CUHI at different time instances during the diurnal cycle have not received sufficient attention. To address these issues, this study employs a relatively large sample of 1,058 microclimate sensors in Seoul, Korea, and analyzes CUHI at 9 a.m., 3 p.m., and 9 p.m. on a typical summer day. Gradient boosting decision tree (GBDT) based regression model and linear regression model and their variants with spatial information were constructed and compared for each time instance. Results show that spatially explicit GBDT models had the highest accuracy. Feature importance analysis shows that built form factors such as mean building height were more important during the afternoon, while surface fractions such as road surface fraction had greater influences during the morning and nighttime. Partial dependence plots (PDPs) show that urban form features influence CUHI in a complex nonlinear manner that varies for each time instance. PDPs were further scrutinized based on their activation and threshold effects. GBDT model findings directionally aligned with linear regression, but they provided nuanced insights into the form-CUHI relationship. These findings help planners to better understand the complexity of urban climate and intervention required to reduce CUHI magnitude across the diurnal cycle.

Predictive geochemical mapping using machine learning in western Kenya

Digital soil mapping techniques represent a cost-effective method for obtaining detailed information regarding the spatial distribution of chemical elements in soils. Machine learning (ML) algorithms using random forest (RF) models have been developed for classification, pattern recognition and regression tasks, they are capable of modelling non-linear relationships using a range of datasets, identifying hierarchical relationships, and determining the importance of predictor variables. In this study, we describe a framework for spatial prediction based on RF modelling where inverse distance weighted (IDW) predictors are used in conjunction with ancillary environmental covariates. The model was applied to predict the total concentration (mg kg−1) and assess the prediction uncertainty of 56 elements, soil pH and organic matter content using 466 soil samples in western Kenya; the results of iodine (I), selenium (Se), zinc (Zn) and soil pH are highlighted in this work. These elements were selected due to contrasting biogeochemical cycles and widespread dietary deficiencies in sub-Saharan Africa, whilst soil pH is an important parameter controlling soil chemical reactions. Algorithm performance was evaluated determining the relative importance of each predictor variable and the model's response using partial dependence profiles. The accuracy and precision of each RF model were assessed by evaluating out-of-bag predicted values. The models R2 values range from 0.31 to 0.64 whilst CCC values range from 0.51 to 0.77. The IDW predictor variables had the greatest impact on assessing the distribution of soil properties in the study area, however, the inclusion of ancillary environmental data improved model performance for all soil properties. The results presented in this paper highlight the benefits of ML algorithms which can incorporate multiple layers of data for spatial prediction, uncertainty assessment and attributing variable importance. Additional research is now required to ensure health practitioners and the agri-community utilise the geochemical maps presented here for assessing the relationship between environmental geochemistry, endemic diseases and preventable micronutrient deficiency.

Enhancing Benzene Combustion Activity through Preferential Platinum Atom Exposure via Strategic Pt-Cu Alloying.

Volatile organic compounds such as benzene are hazardous air pollutants that require effective elimination. Noble metal-based catalysts exhibit high benzene combustion activity, but their prohibitive cost necessitates strategies to enhance utilization efficiency. This study investigates a Pt-Cu alloy catalyst for improved benzene combustion by preferentially exposing Pt active sites through Cu alloying. Aberration-corrected scanning transmission electron microscopy and X-ray spectroscopy characterize the nanoscale distribution and enrichment of Pt on the alloy surface. Kinetic measurements demonstrate substantially enhanced activity compared with Pt catalysts, attributed to increased Pt metallic site exposure rather than alteration of the reaction mechanism. In situ Fourier transform infrared (FTIR) spectroscopy reveals a higher abundance of terrace-like Pt sites in the alloy, beneficial for benzene adsorption. Partial pressure dependence analyses indicate competitive adsorption of benzene and O2, following Langmuir-Hinshelwood kinetics. These findings provide conceptual insights into tuning surface composition in bimetallic catalysts to optimize noble metal efficiency, with broad applicability for sustainable catalytic process advancement.

Analysing the determinants of the Turkish household electricity consumption using gradient boosting regression tree

Residential buildings are the second largest electricity consumer in Turkey. Thus, the goal here is to detect the factors that determine the electricity consumption of the households in Turkey using the Household Budget Survey (HBS). This study applies Decision Tree (DT), Random Forest (RF) and Gradient Boosted Regression Tree (GBRT) methods. Since the GBRT method provides the lowest Root Mean Squared Error (RMSE), the impact of each variable on the electricity consumption is analysed with this method. The most critical determinant is found to be the household size, while income level and heating type are discovered as 2nd and 3rd most prominent determinants for household electricity demand. With the help of the Partial Dependence Plots (PDP) provided by the GBRT method, the impact of each categorical and continuous variable is presented. Using the results of PDPs, the monetary values of both electricity generation and the social cost of CO2 emissions emitted into the atmosphere due to electricity generation are calculated for the most important determinants.

Cross‐entropy‐based adaptive fuzzy control for visual tracking of road cracks with unmanned mobile robot

AbstractVisual tracking of road cracks in unstructured road environment was, is, and remains a crucial and challenging task, which plays a vital role in accurate crack sealing for automated road cracks repair. However, many problems have not been well solved in existing automated road cracks repair, such as the low automation due to partial dependence on manual and the interrupted traffic flow caused by the heavy equipment used. In this article, a cross‐entropy‐based adaptive fuzzy control (CEAFC) method is proposed, which reaches visual tracking with unmanned mobile robot (VT‐UMbot) for road cracks. Specifically, the CEAFC method uses cross‐entropy optimization iteration to tune parameters for the tracking controller, and fuzzy logic is constructed to explore robustness improvement. Moreover, a framework of VT‐UMbot based on a four‐wheel independent differential drive is established, and visual servo and tracking control are integrated into the system. Our experiment shows that the proposed method is extensively evaluated on three road cracks scenarios and achieves state‐of‐the‐art performance with high efficiency.

Identifying the drivers of chlorophyll-a dynamics in a landscape lake recharged by reclaimed water using interpretable machine learning

The water quality of lakes recharged by reclaimed water is affected by both the fluctuation of reclaimed water quality and the biochemical processes in the lakes, and therefore the main controlling factors of algal blooms are difficult to identify. Taking a typical landscape lake recharged by reclaimed water as an example and using the spatiotemporal distribution characteristics and correlation analysis of water quality indexes, we propose an interpretable machine learning framework based on random forest to predict chlorophyll-a (Chl-a). The model considered nutrient difference indexes between reclaimed water and lake water, and further used feature importance ranking and partial dependence plot to identify nutrient drivers. Results show that the NO3−-N input from reclaimed water is the dominant nutrient driver for algal bloom especially at high temperatures, and the negative correlation between NO3−-N and Chl-a in the lake water is the consequence of algal bloom rather than the cause. Our study provides new insights into the identification of eutrophication factors for lakes recharged by reclaimed water.

Physicochemical analysis of infiltrated cathode symmetric cells for intermediate temperature SOFCs

This study investigated the cathode parameter changes before and after infiltration based on Jamnik and Maier's MIEC transmission line model. To consider additional factors such as surface adsorption capacitance, chemical capacitance Cchem, and general capacitance effect CHN were introduced in parallel. To obtain the appropriate temperature dependence, oxygen partial pressure dependence, and current density dependence trends of transmission line parameters, the trend of Cchem was fixed according to each measurement condition using the thermodynamic parameters of previous papers. Infiltrating Sm0.5Sr0.5CoO3-δ (SSC), a better catalyst material, into the widely used La0.6Sr0.4Co0.2Fe0.8O3-δ (LSCF6428) perovskite electrode resulted in a decrease in surface reaction resistance Rp by about 1 order of magnitude and a decrease in oxygen ion diffusion resistance Rs. The oxygen ion adsorption capacitance CHN0 on the surface is larger after infiltration due to the larger surface area of the infiltrated electrode, which can be explained by the two arcs clearly separated in the impedance spectra of the infiltrated cell. Additionally, after infiltration, the diffusion characteristics depending on oxygen partial pressure changed relatively from bulk diffusion to surface diffusion. The value of Rs is larger than Rp under almost all conditions, indicating that the overall rate-determining step (rds) is the diffusion reaction in the current direction. Furthermore, to observe the changes in surface reaction characteristics only, the characteristics of Rp before and after infiltration were analyzed based on the Butler-Volmer-type equation and oxygen partial pressure dependence of the exchange current density of surface reactions. As a result, the rds of surface reactions changed from the reduction reaction of the intermediate stage of electrode reactions after infiltration to the dissociative adsorption reaction of the initial reaction stage.

Machine learning approach for investigating compressive strength of self-compacting concrete containing supplementary cementitious materials and recycled aggregate

Supplementary cementitious materials (SCMs) and recycled coarse aggregate (RCA) have the potential for sustainable development and resource utilization and have been widely applied in self-compacting concrete (SCC). However, traditional experiment approaches for predicting its compressive strength suffer from inefficiency, time-consuming, and high cost. To address these issues, machine learning (ML) models are employed, evaluated, and compared in this paper. First, a database of 337 samples with 10 features is collected and established. Then, the 12 ML models, including SVM, KNN, DT, RF, ANN, GBoost, XGBoost, AdaBoost, CatBoost, LightGBM, HistGBM, and GEP are applied for modeling to predict the compressive strength. Finally, both feature importance, individual conditional expectation, and partial dependence plot are utilized to assess the importance and the effect of each and coupled parameters. The results show that the CatBoost and LightGBM models generally possess superior performance in various aspects, including higher R2 values, lower RMSE, MSE, and MAE values, as well as smaller errors. Among the Boosting models, except for the AdaBoost model, satisfactory predictive ability is observed. The ANN and RF models also demonstrate relatively good performance, and the SVM, DT, KNN, and GEP models present relatively poorer performance. Curing age and W/CMs exhibit the most pivotal features on model prediction. Furthermore, RCA replacement ratio plays a more prominent role than CA, and SCMs contribute somewhat to the compressive strength of concrete, but not as much compared to cement. Also, it is important to acknowledge the significant contribution of SP. This study offers a systematic evaluation of the prediction of compressive strength in SCC containing SCMs and RCA, making a notable contribution to both the existing literature and practical applications.

Machine learning prediction of biochar yield based on biomass characteristics

Slow pyrolysis is a widely used thermochemical pathway that can convert organic waste into biochar. We employed six machine learning models to predictively model 13 selected variables using pearson feature selection. Additionally, partial dependence analysis is used to reveal the deep relationship between feature variables. Both the gradient boosting decision tree and the Levenberg-Marquardt backpropagation neural network achieved training set R2 > 0.9 and testing set R2 > 0.8. But the other models displayed lower performance on the testing set, with R2 < 0.8. The partial dependence plot demonstrates that pyrolysis conditions have greater impact on biochar yield than biomass composition. Furthermore, the highest treatment temperature, being the sole consistently changing feature, can serve as a guiding factor for regulating biochar yield. This study highlights the immense potential of machine learning in experimental prediction, providing a scientific reference for reducing time and economic costs in pyrolysis experiments and process development.

Machine learning prediction of bio-oil yield during solvothermal liquefaction of lignocellulosic biowaste

Solvothermal liquefaction of biomass has gained attention to produce liquid biofuel and specialized chemicals. In this study, eXtreme Gradient Boosting was applied for predicting the bio-oil yield during solvothermal liquefaction of lignocellulosic biowaste. To establish a precise model for predicting the bio-oil yield, the prediction of the biomass conversion was found to be an intermediate ameliorating variable. The combination of the contents of biochemical components (cellulose, hemicellulose and lignin) with the operating factors (temperature, time, solid loading, solvent polarity and solvent density) provided the best prediction for the biomass conversion (R2 equals to 99.98% for training and 97.67% for test). To predict the yield of bio-oil, introduction of the biomass conversion among inputs improved prediction accuracy (R2 equals to 100% for training and 94.4% for test). The best prediction models developed were interpreted using a game theory-based feature importance and the partial dependence plotting analysis, which provided insights into the biomass conversion pathways and the bio-oil generation mechanism. To assist other researchers in predicting, a graphical user interface was created. This tool will save both resources and time that would otherwise be spent on multiple experimental trials and might not yield useful results.

Assessing the weaponability of enriched uranium trade in the geopolitics of nuclear energy: The EU-Russia interrelations

Russia's invasion of Ukraine in 2022 revived the debates over Moscow's grip on the nuclear fuel market. European and American governments had since planned for policies to diminish their reliance on Russian supplies, motivated by fear of the Kremlin using it as a coercive diplomatic tool, drawing parallels with experiences of fossil fuels exportation disruptions. While global interest for atomic energy is on the rise again, this article aims at assessing the weaponability of nuclear fuel trades, i.e., the conditions in which they can be used by suppliers to politically constrain importing nations. Geopolitics scholarship had until recently paid little attention to nuclear power and the few existing works overlooked the supply issue. In this paper, we bridge existing literature on the “energy weapon” to build a renewed analytical framework to identify the conditions in which fuel trade interdependencies can be coercively used. We then applied it to the case of enriched uranium trade between Russia and European Union's member states. Data were gathered through analysis of a textual corpus consisting of articles from three main magazines focused on nuclear power—Nuclear Engineering International, World Nuclear News and NucNet — (n = 439), 14 semi-structured interviews and three international conferences observations. Results show that Russia's weaponization capacity of enriched uranium trades against Europe in low, but not inexistent, thanks to supply diversification strategies, existing inventories, ongoing production overcapacity, transport fluidity and partial dependence of Rosatom on exports. However, countries' vulnerability isn't homogenous through Europe, and the low weapon ability of nuclear fuel is largely dependent on the fragile equilibrium in the supply and demand balance. Finally, this paper identifies the limits of applying weaponization theories developed in the context of fossil fuel trades to the case of nuclear power.

머신러닝을 활용한 유치원 정보공시 만족도 영향 요인 분석

The purpose of this study is to identify the use of kindergarten notifications, which can be the realization of the information disclosure system and to predict overall satisfaction using statistical analysis methods and random forest algorithms. Based on the results of the annual kindergarten notification satisfaction survey conducted by the Korea Educational Information Service, differences in user response were analyzed, and random forest was used to identify factors that affect overall satisfaction prediction among machine learning algorithms. As a result of the t-test and one-way distribution variance analysis, the difference in use status according to gender, age, residential area, and type was statistically significant. As a result of random forest model learning, Accuracy 0.85 and F1 Score 0.84 were shown. Explanable artificial intelligence such as variable importance, SHAP plot, and partial dependence charts of the learned model was used to identify variables that affect the prediction of kindergarten information disclosure. It was confirmed that the usage status variable contributed more to the satisfaction prediction than the user's background variable, and based on the results, the direction of improvement of kindergarten information disclosure and kindergarten notification was derived.

Applied machine learning to analyze and predict CO2 adsorption behavior of metal-organic frameworks

Machine learning provides new insights for designing MOFs with high CO2 adsorption capacity and understanding the CO2 adsorption mechanism. In this work, 348 data points from published reports were collected and four tree-based models were designed to predict the CO2 adsorption capacity of MOFs by machine learning. The results showed that the Random Forest (RF) had the best prediction performance (R2train = 0.970, R2test = 0.896). Feature importance analysis revealed the relative importance of CO2 adsorption parameters (73 %), textures (23 %) and metal centers of MOFs (4 %) for the CO2 adsorption process. Single and synergistic effects of different features were observed through partial dependence analysis. MOFs with Cu, Fe, Co, and Ni metal centers exhibited a promoting effect on CO2 adsorption. In addition, under high pressure, well-developed textures had significant positive impact on CO2 adsorption capacity, while under medium and low pressure, textures were not determining factors.

In Vitro Antioxidant and In Vivo Antigenotoxic Features of a Series of 61 Essential Oils and Quantitative Composition-Activity Relationships Modeled through Machine Learning Algorithms.

The antioxidant activity of essential oils (EOs) is an important and frequently studied property, yet it is not sufficiently understood in terms of the contribution of EOs mixtures' constituents and biological properties. In this study, a series of 61 commercial EOs were first evaluated as antioxidants in vitro, following as closely as possible the cellular pathways of reactive oxygen species (ROS) generation. Hence, EOs were assessed for the ability either to chelate metal ions, thus interfering with ROS generation within the respiratory chain, or to neutralize 2,2-diphenyl-1-picrylhydrazyl (DPPH•) and lipid peroxide radicals (LOO•), thereby halting lipid peroxidation, as well as to neutralize 2,2'-azino-bis(3-ethylbenzothiazoline-6-sulfonic acid cation radicals (ABTS•+) and hydroxyl radicals (OH•), thereby preventing the ROS species from damaging DNA nucleotides. Showing noteworthy potencies to neutralize all of the radicals at the ng/mL level, the active EOs were also characterized as protectors of DNA double strands from damage induced by peroxyl radicals (ROO•), emerging from 2,2'-azobis-2-methyl-propanimidamide (AAPH) as a source, and OH•, indicating some genome protectivity and antigenotoxicity effectiveness in vitro. The chemical compositions of the EOs associated with the obtained activities were then analyzed by means of machine learning (ML) classification algorithms to generate quantitative composition-activity relationships (QCARs) models (models published in the AI4EssOil database available online). The QCARs models enabled us to highlight the key features (EOSs' chemical compounds) for exerting the redox potencies and to define the partial dependencies of the features, viz. percentages in the mixture required to exert a given potency. The ML-based models explained either the positive or negative contribution of the most important chemical components: limonene, linalool, carvacrol, eucalyptol, α-pinene, thymol, caryophyllene, p-cymene, eugenol, and chrysanthone. Finally, the most potent EOs in vitro, Ylang-ylang (Cananga odorata (Lam.)) and Ceylon cinnamon peel (Cinnamomum verum J. Presl), were promptly administered in vivo to evaluate the rescue ability against redox damage caused by CCl4, thereby verifying their antioxidant and antigenotoxic properties either in the liver or in the kidney.

Dinaciclib synergizes with BH3 mimetics targeting BCL-2 and BCL-XL in multiple myeloma cell lines partially dependent on MCL-1 and in plasma cells from patients.

A better understanding of multiple myeloma (MM) biology has led to the development of novel therapies. However, MM is still an incurable disease and new pharmacological strategies are needed. Dinaciclib, a multiple cyclin-dependent kinase (CDK) inhibitor, which inhibits CDK1, 2, 5 and 9, displays significant antimyeloma activity as found in phase II clinical trials. In this study, we have explored the mechanism of dinaciclib-induced death and evaluated its enhancement by different BH3 mimetics in MM cell lines as well as in plasma cells from MM patients. Our results indicate a synergistic effect of dinaciclib-based combinations with B-cell lymphoma 2 or B-cell lymphoma extra-large inhibitors, especially in MM cell lines with partial dependence on myeloid cell leukemia sequence 1 (MCL-1). Simultaneous treatment with dinaciclib and BH3 mimetics ABT-199 or A-1155463 additionally showed a synergistic effect in plasma cells from MM patients, ex vivo. Altered MM cytogenetics did not affect dinaciclib response ex vivo, alone or in combined treatment, suggesting that these combinations could be a suitable therapeutic option for patients bearing cytogenetic alterations and poor prognosis. This work also opens the possibility to explore cyclin-dependent kinase 9 inhibition as a targeted therapy in MM patients overexpressing or with high dependence on MCL-1.

Explainable Artificial Intelligence in Hydrology: Interpreting Black-Box Snowmelt-Driven Streamflow Predictions in an Arid Andean Basin of North-Central Chile

In recent years, a new discipline known as Explainable Artificial Intelligence (XAI) has emerged, which has followed the growing trend experienced by Artificial Intelligence over the last decades. There are, however, important gaps in the adoption of XAI in hydrology research, in terms of application studies in the southern hemisphere, or in studies associated with snowmelt-driven streamflow prediction in arid regions, to mention a few. This paper seeks to contribute to filling these knowledge gaps through the application of XAI techniques in snowmelt-driven streamflow prediction in a basin located in the arid region of north-central Chile in South America. For this, two prediction models were built using the Random Forest algorithm, for one and four months in advance. The models show good prediction performance in the training set for one (RMSE:1.33, R2: 0.94, MAE:0.55) and four (RMSE: 5.67, R2:0.94, MAE: 1.51) months in advance. The selected interpretation techniques (importance of the variable, partial dependence plot, accumulated local effects plot, Shapley values and local interpretable model-agnostic explanations) show that hydrometeorological variables in the vicinity of the basin are more important than climate variables and this occurs both for the dataset level and for the months with the lowest streamflow records. The importance of the XAI approach adopted in this study is discussed in terms of its contribution to the understanding of hydrological processes, as well as its role in high-stakes decision-making.