Model Skill Research Articles

Extreme Events Contributing to Tipping Elements and Tipping Points

AbstractThis review article provides a synthesis and perspective on how weather and climate extreme events can play a role in influencing tipping elements and triggering tipping points in the Earth System. An example of a potential critical global tipping point, induced by climate extremes in an increasingly warmer climate, is Amazon rainforest dieback that could be driven by regional increases in droughts and exacerbated by fires, in addition to deforestation. A tipping element associated with the boreal forest might also be vulnerable to heat, drought and fire. An oceanic example is the potential collapse of the Atlantic meridional overturning circulation due to extreme variability in freshwater inputs, while marine heatwaves and high acidity extremes can lead to coral reef collapse. Extreme heat events may furthermore play an important role in ice sheet, glacier and permafrost stability. Regional severe extreme events could also lead to tipping in ecosystems, as well as in human systems, in response to climate drivers. However, substantial scientific uncertainty remains on mechanistic links between extreme events and tipping points. Earth observations are of high relevance to evaluate and constrain those links between extreme events and tipping elements, by determining conditions leading to delayed recovery with a potential for tipping in the atmosphere, on land, in vegetation, and in the ocean. In the subsurface ocean, there is a lack of consistent, synoptic and high frequency observations of changes in both ocean physics and biogeochemistry. This review article shows the importance of considering the interface between extreme events and tipping points, two topics usually addressed in isolation, and the need for continued monitoring to observe early warning signs and to evaluate Earth system response to extreme events as well as improving model skill in simulating extremes, compound extremes and tipping elements.

Balloon Baseline Stratospheric Aerosol Profiles (B2SAP)—Perturbations in the Southern Hemisphere, 2019–2022

AbstractVolcanic and pyrocumulonimbus (pyroCB) injections into the stratosphere perturb the aerosol layer and can have important radiative and chemical impacts on timescales spanning from months to several years. Repeated in situ balloon‐borne measurements of aerosol size and number concentration (>140 nm in diameter), ozone, water vapor, and atmospheric state variables made at midlatitudes in the southern hemisphere (SH) since 2019 enable us to better characterize such events. We use this record and coincident lidar extinction profiles to study several moderate to large stratospheric perturbations in the SH between 2019 and 2022 in detail, including the Australian New Year Super Outbreak (ANYSO) pyroCB in 2020. Median vertical profiles of aerosol number concentration, effective radius, and surface area in SH midlatitudes are also compared with those recorded in Northern Hemisphere midlatitudes under baseline conditions using an identical payload. These data depict the variability in stratospheric aerosol properties in the SH midlatitudes during this period and provide a benchmark for global sectional aerosol models. They reveal that sulfate particle size distributions under baseline conditions and in volcanic plumes are relatively well represented in the Community Earth System Model—Community Aerosol Radiation Model for Atmospheres (CESM‐CARMA), but more observations of biomass burning plumes are needed to improve model skill in simulating pyroCB. Comparisons between in situ and lidar observations also highlight a need for more observations of aerosol composition and refractive index in both fresh and aging biomass burning plumes.

Pengaruh Model Pembelajaran Discovery Learning dan Keterampilan Berpikir Kreatif Terhadap Pemahaman Konsep Biologi Siswa SMA Bunda Hati Kudus

In the era of globalization marked by the presence of Society 5.0, education faces new challenges. One of the essential skills that students need to have is critical thinking. The discovery learning learning model is considered effective in fostering students' critical thinking skills, so that it can improve their understanding of concepts. This study aims to determine the influence of the discovery learning model and creative thinking skills on the understanding of biology concepts of SMA Bunda Hati Kudus students. The research method used is a quantitative method. The population of this study used 160 students in class XI science of Bunda Hati Kudus High School for the 2022/2023 school year. The sample used was 64 people. Hypothesis testing using MANOVA and ANOVA techniques with a significance level of 0.05. The results of the study show that the application of the discovery learning learning model significantly improves the understanding of concepts, critical thinking skills, and creative thinking skills of SMA Bunda Hati Kudus students. Statistical analysis with the F test showed a very significant difference (p < 0.05) between the group of students who used the discovery learning model and the group that used the direct learning model. These findings indicate that the discovery learning model can be an effective alternative to improve the quality of student learning

A Framework for Using GenAI to Support Student Engagement in Interdisciplinary Learning from Self-determination Theory

Artificial intelligence (AI) is an interdisciplinary field, including mathematics, sciences, psychology, linguistics, sociology, and engineering (Bisconti et al., 2023; Chen, 2020; Zhuang et al., 2020). The development of its technologies requires experts from various disciplines. In education, generative AI (GenAI) tools such as ChatGPT and Dall-E facilitate interdisciplinary learning by breaking down barriers between disciplines and enabling students to see connections across different disciplines, resulting in a more holistic understanding of complex concepts. The tools enable students from different disciplines to collaborate on projects by combining their diverse knowledge and skills (Chiu, 2023, 2024; Chiu et al., 2023). For example, in STEAM education, students who are weak at coding can develop simple applications by asking GitHub Copilot, which generates specific code suggestions including variables, classes, and methods based on their situations. Those who are poor in art designs, can ask t Dall-E to generate visuals, fostering an interdisciplinary approach. Therefore, students can use the tools to complement the knowledge gaps they have because they have instant access to information and resources related to other disciplines (Kusters et al., 2020; Peters et al., 2023). Students become more adaptable, open, flexible, and innovative in their thinking. These enhance critical thinking skills, creativity, and problem-solving skills in students. This interdisciplinary learning using AI also fosters a sense of curiosity and a willingness to explore new ideas, which are essential skills for success in the AI era, better preparing students to address real-world problems and excel in a variety of disciplines. Most related current studies focus on how GenAI affects assessment and essay-based assignments (Chiu, 2024; Xia et al., 2024). However, student engagement in using GenAI tools in completing interdisciplinary learning remains unclear. Student engagement is a multi-dimensional concept generally considered to include behavioral, emotional, cognitive, and agentic components (Chiu, 2021; Fredricks, 2011), which can be explained by self-determination theory (SDT) through needs satisfaction (autonomy: feel having choices; competence: feel capable; relatedness: feel relevant) (Chiu, 2021; Ryan and Deci, 2020). This mixed-methods study investigated how GenAI tools engage students in interdisciplinary learning from SDT perspectives. The two main research questions are: RQ1: How is needs satisfaction related to student engagement in interdisciplinary learning? RQ2: How do GenAI tools satisfy students three SDT needs?. The participants were 310 university students from various academic major backgrounds (sciences: 72; engineering: 78; social sciences: 79; art: 76; others: five). All of them attended a 3-hour workshop on how to use GenAI in completing a project under the topic of sustainability. The participants tried different GenAI tools, and discussed how they completed the project. A questionnaire with validated itemswas used to examine how GenAI tools satisfy their needs for competency in a sustainability project (RQ1) (Chiu, 2024). Three iterative cycles of group interviews were conducted with randomly selected 30 participants to understand the roles and affordances of GenAI in interdisciplinary learning (RQ2). The quantitative findings suggested that during interdisciplinary learning, GenAI can satisfy student needs by fostering classification and generalization skills (grouping knowledge), encouraging unpacking model and solution skills (revising thinking), and providing alternative intelligence (inspiring with new ideas). The qualitative findings from this study were used to propose a framework for using GenAI to support interdisciplinary learning. The framework also visualizes the roles and affordances of GenAI in the learning.

The Effect of Problem Based Learning Models and Analysis Skills on Mathematics Learning Outcomes in Primary Schools

This research aims to determine the effect of the problem based learning model and analytical skills on elementary school students' mathematics learning outcomes. This type of research is quantitative research using experimental methods with a research design using a 2x2 factorial design. The research was carried out at SDN 3 Lenek Daya. Data collection techniques use tests to determine learning outcomes and students' level of analysis. The results of the research show that 1) there is a difference in the mathematics learning outcomes of students who are taught using the Problem Based Learning (PBL) and the learning outcomes of students who use the Cooperative Think-Pair-Share (TPS) model, indicated by the difference in Fhitung values ​​greater than Ftable by (17,991 > 3.44). 2) there is an interaction effect between the PBL and analytical skills on students' mathematics learning outcomes, as evidenced by the Fhitung value being greater than Ftable (4.549 > 3.44) with a Sig value < α value (0.029 < 0.05). 3) there are differences in mathematics learning outcomes for students who have a high analysis of learning using PBL and TPS, this is because Fhitung is greater than Ftable (6.472 > 3.44), as well as the sig value < α value (0.030 < 0.05). 4) there are differences in mathematics learning outcomes for students who have low analytical skills who are taught using the PBL and TPS models, because the Fhitung value is greater than Ftable (19,950 > 3.44) while the sig value < α value (0.001 < 0.05)

Systematic Upstream Large-Scale Control of Subtropical Low-Cloud Properties

Abstract It has recently been shown that the slow adjustment of the atmospheric boundary layer (ABL) to perturbations of the large-scale atmospheric conditions translate into an upstream control of the low-cloud cover (LCC) variability at climatological timescales in the subtropics. In this study we expand upon this recent study to investigate upstream control of the climatology of low-cloud radiative effect (CRE), as well as cloud properties relevant to their radiative effect: cloud liquid water path (LWP), and cloud optical depth (COD).We use machine-learning statistical models with feature selection capabilities (random-forests) to determine the influence of the local and upstream large-scale conditions in monthly data. These conditions are determined using back-trajectories in monthly-mean wind fields. Total CRE, dominated by the shortwave contribution, exhibits a dependence on upstream Estimated Inversion Strength (EIS), but not on upstream Sea-Surface Temperature (SST) as LCC does. Upstream control of COD was present but was not as consistent as LCC or CRE across different regions and cloud regimes, while LWP does not exhibit strong upstream control at all. This implies that the control of other boundary layer properties could explain the differences in response between LCC and CRE to SST and EIS. Looking at five subtropical eastern basins, it appears that key lead times are region-specific. The inclusion of upstream control provides significant improvements to the predictive skill of statistical models over local linear regression, with improvements in variance explained well over 20% in many cases.

Онтологическое моделирование как средство семантизации терминологической лексики в процессе обучения иностранному языку специальности

The study provides a rationale for the use of ontological modeling techniques for teaching the language of a specialty to foreign students studying educational programs in Russian. Ontologies of subject areas, defined as conceptualization specifications, are widely used in computational linguistics to create formalized knowledge management systems. In the context of internationalization, globalization and a colossal increase in the volume of industry knowledge, linguaprofessional skills of interlingual modeling and systematization of conceptual areas of various professions should become a basic element of foreign-language professional communicative competence. This study presents the experience of applying linguistic ontological modeling techniques as a means of improving the effectiveness of terminology teaching, in particular, in terms of semantization of terminological units that form the conceptual sphere of professional communication of students in non-linguistic areas of training. In the course of this study, positive experience was accumulated in applying linguistic ontological modeling techniques, proving the feasibility of using ontologies as a means of semantization of industry terminology. In the applied aspect, the approach we propose can be used for ontological modeling of foreign language courses for special purposes, as well as for creating digital didactic products with an extended list of functions that form a linguistic and professional virtual learning environment.

Improving subseasonal forecast skill in the Norwegian Climate Prediction Model using soil moisture data assimilation

AbstractThis study shows the importance of soil moisture (SM) in subseasonal-to-seasonal (S2S) predictions at mid-latitudes. We do this through introducing the Norwegian Climate Prediction Model Land (NorCPM-Land), a land reanalysis framework tailored for integration with the Norwegian Climate Prediction Model (NorCPM). NorCPM-Land assimilates blended SM data from the European Space Agency’s Climate Change Initiative into a 30-member offline simulation of the Community Land Model with fluxes from the coupled model. The assimilation of SM data reduces error in SM by 10.5 % when validated against independent SM observations. It also improves latent heat flux estimates, illustrating that the adjustment of underlying SM significantly augments the capacity to model land surface dynamics. We evaluate the added value of land initialisation for subseasonal predictions, by comparing the performance of hindcasts (retrospective prediction) using the standard NorCPM with a version where the land initial condition is taken from NorCPM-Land reanalysis. The hindcast covers the period 2000 to 2019 with four start dates per year. Land initialisation enhances SM predictions, reducing error by up to 2.5-month lead time. Likewise, the error for precipitation and temperature shows improvement up to a lead time of 1.5-month. The largest improvements are observed in regions with significant land-atmospheric coupling, such as the Central United States, the Sahel, and Central India. This method further enhances the prediction of extreme temperature variations, both high and low, with the most notable improvements seen in regions at mid and high latitudes, including parts of Europe, the United States, and Asia. Overall, our study provides further evidence for the significant role of SM content in enhancing the accuracy of subseasonal predictions. This study introduces a technique for improved land initialisation, utilising the same model employed in climate predictions.

Standardized Daily High‐Resolution Large‐Eddy Simulations of the Arctic Boundary Layer and Clouds During the Complete MOSAiC Drift

AbstractThis study utilizes the wealth of observational data collected during the recent Multidisciplinary drifting Observatory for the Study of Arctic Climate (MOSAiC) drift experiment to constrain and evaluate close to two‐hundred daily Large‐Eddy Simulations (LES) of Arctic boundary layers and clouds at high resolutions. A standardized approach is adopted to tightly integrate field measurements into the experimental configuration. Covering the full drift represents a step forward from single‐case LES studies, and allows for a robust assessment of model performance against independent data under a range of atmospheric conditions. A homogeneously forced domain is simulated in a Lagrangian frame of reference, initialized with radiosonde and value‐added cloud profiles. Prescribed boundary conditions include various measured surface characteristics. Time‐constant composite forcing is applied, primarily consisting of subsidence rates sampled from reanalysis data. The simulations run for 3 hours, allowing turbulence and clouds to spin up while still facilitating direct comparison to MOSAiC data. Key aspects such as the vertical thermodynamic structure, cloud properties, and surface energy fluxes are well reproduced and maintained. The model captures the bimodal distribution of atmospheric states that is typical of Arctic climate. Selected days are investigated more closely to assess the model's skill in maintaining the observed boundary layer structure. The sensitivity to various aspects of the experimental configuration and model physics is tested. The model input and output are available to the scientific community, supplementing the MOSAiC data archive. The close agreement with observed meteorology justifies the use of LES for gaining further insight into Arctic boundary layer processes and their role in Arctic climate change.

The effect of climate change and temperature extremes on Aedes albopictus populations: a regional case study for Italy.

The Asian tiger mosquito, Aedes albopictus, has spread widely throughout Italy since its introduction, with significant public health implications. We examine how decadal temperature trends and sub-monthly heatwave events affect its climate-driven geographical distribution and temporal dynamics using a new regional-scale dynamical Aedes model. The model is calibrated using [Formula: see text] years of ovitrap data for Emilia-Romagna and reproduces the vector seasonality and, to a lesser extent, its inter-annual variability. Simulated vector density hotspots overlap with densely populated areas in Rome, Milan, Naples, Foggia, Catania, Palermo, Lecce, Cagliari, Genoa, Turin and large urban centres in Emilia-Romagna. Lower risk is simulated over the Central Apennine mountains and the Alps. At decadal time scale, we simulate a lengthening of the active mosquito season by 0.5-3 weeks per decade, with the vector becoming homodynamic in southern Italy. Depending on the climatic setting, heatwaves can increase or reduce vector populations and, in some locations, can temporarily decrease mosquito populations. Such decreases can be followed by a population rebound and overshoot. Given the model's skill in reproducing key spatio-temporal Ae. albopictus features, there is potential to develop an early warning system to inform control efforts at a national scale.

Verification of Operational Forecast Models in Cases of Extratropical Transition of North Atlantic Hurricanes

Abstract Operational forecast models are necessary for the prediction of weather events in real time. Verification of these models must be performed to assess model skills and areas in need of improvement, particularly with different types of weather events that may occur. Despite the devastating impacts that can be caused by tropical cyclones (TCs) that undergo extratropical transition (ET) and become post-tropical cyclones (PTCs), these storms have not been extensively studied in the context of short-term weather prediction. This study completes the first analysis of the Global Forecast System (GFS) and a preoperational version of the newly operational Hurricane Analysis and Forecast System (HAFS) models in forecasting the occurrence of ET and the rainfall associated with ET storms in the North Atlantic basin. GFS’s skill exceeds that of HAFS in forecasting the occurrence of ET, but HAFS tends to have lower track and rain-rate errors in the fully tropical phase of ET storms’ life cycles. Both models simulate rain rates that are often too high near the storm center and fail to capture the larger area of moderate rain rates that greatly contributes to total rainfall accumulation. The discrepancies in rain rates between the models and Integrated Multi-satellitE Retrievals for GPM (IMERG) could be attributed to the models’ tendency to keep storms too intense and too compact with an overly strong warm core, even throughout the ET process.

Impact‐Based Skill Evaluation of Seasonal Precipitation Forecasts

AbstractWe introduce an impact‐based framework to evaluate seasonal forecast model skill in capturing extreme weather and climate events over regions prone to natural disasters such as floods and wildfires. Forecasting hydroclimatic extremes holds significant importance in an era of increasing hazards such as wildfires, floods, and droughts. We evaluate the performance of five Copernicus Climate Change Service (C3S) seasonal forecast models (CMCC, DWD, ECCC, UK‐Met, and Météo‐France) in predicting extreme precipitation events from 1993 to 2016 using 14 indices reflecting timing and intensity (using absolute and locally defined thresholds) of precipitation at a seasonal timescale. Performance metrics, including Percent Bias, Kendall Tau Rank Correlation Score, and models' discrimination capacity, are used for skill evaluation. Our findings indicate that the performance of models varies markedly across regions and seasons. While models generally show good skill in the tropical regions, their skill in extra‐tropical regions is markedly lower. Elevated precipitation thresholds (i.e., higher intensity indices) correlate with heightened model biases, indicating deficiencies in modeling severe precipitation events. Our analysis using an impact‐based framework highlights the superior predictive capabilities of the UK‐Met and Météo‐France models in capturing the underlying processes that drive precipitation events, or lack thereof, across many regions and seasons. Other models exhibit strong performance in specific regions and/or seasons, but not globally. These results advance our understanding of an impact‐based framework in capturing a broad spectrum of extreme weather and climatic events, and inform strategic amalgamation of diverse models across different regions and seasons, thereby offering valuable insights for disaster management and risk analysis.

Developing Workforce with Mathematical Modeling Skills

Developing Workforce with Mathematical Modeling Skills

New trends in didactic research in university mathematics education

AbstractRecent research in university mathematics education has moved beyond the traditional focus on the transition from secondary to tertiary education and students' understanding of introductory courses such as pre-calculus and calculus. There is growing interest in the challenges students face as they move into more advanced mathematics courses that require a shift toward formal reasoning, proof, modeling, and problem-solving skills. This survey paper explores emerging trends and innovations in the field, focusing on three key areas: innovations in teaching and learning advanced mathematical topics, transitions between different levels and contexts of mathematics education, and the role of proof and proving in advanced university mathematics. The survey reflects the evolving landscape of mathematics education research and addresses the theoretical and practical challenges of teaching and learning advanced mathematics across various contexts.

Aeolus Winds Improve Arctic Weather Prediction

AbstractIt has been proven that assimilating winds from the Aeolus global Doppler Wind Lidar (DWL) would enhance the predictive skill of weather forecast models. In this study, we use a series of observing system experiments to examine how currently assimilated “operational” wind products and Aeolus winds impact environment and climate change Canada's global forecast system over the data‐sparse Arctic region. Aeolus winds improve the tropospheric wind and temperature forecasts by about 0.7%–0.9% of error reduction (a 15%–20% effect compared to the impact of operational wind products) while having little impact on specific humidity. In particular, Aeolus winds have an impact on forecasts of strong wind days on the wind and temperature fields that is double the impact for weaker wind days. On the other hand, the operational wind products do not show as large a discrepancy in impact between forecasts of disturbed and normal days. These findings suggest significant potential for global DWL observations to enhance severe weather prediction in polar regions.

An empirical study of feature extraction for Hubei folk art of carved paper-cutting.

Excellent traditional culture is the root and soul of a nation, which is the carrier of historical precipitation and continuous development. In this aspect, Hubei folk art of carved paper-cutting has important cultural value and historical significance in the composition of Chinese traditional culture, which is a national intangible cultural heritage under key protection and urgently needs to be carried forward. This paper carried out an empirical study using the survey and comparative analysis methods on the feature extraction for Hubei folk art of carved paper-cutting. Through investigation and comparative analysis of vast artworks of Hubei carved paper-cutting, the classics are selected as the representative to extract the regional characteristics, the artistic values, and the composition and modelling methods. Results indicate that the common features exist on the artistic values in beautification of life and education, the composition principles of central-based, symmetry, balanced or conformal, and the modelling techniques using Yin and Yang carving, unobstructive perspective, distortion and exaggeration. However, the regional features in cultural expression, folklore communication, modelling skills are apparent and strong interlinked with the regional historical development in social activities, labor productivity and surrounding environment. This study provides a multi-dimensional perspective for the modern inheritance and development of traditional Hubei artworks of carved paper-cutting with specified characteristics.

DiaBetter Together: Clinical trial protocol for a strengths-based Peer Mentor intervention for young adults with type 1 diabetes transitioning to adult care

BackgroundType 1 diabetes (T1D) management is challenging for young adults, who are expected to transfer from the pediatric to adult T1D healthcare system while also managing typical developmental demands (e.g., social, financial, work/school, residential). Many young adults have extended gaps in care before following up in adult care, increasing risk for poor health outcomes. There are few evidence-based programs to support young adults with T1D to promote a timelier transition during this period. This paper reports on the design of DiaBetter Together, a randomized controlled trial to evaluate a 12-month Peer Mentor-delivered intervention compared to usual care among young adults with T1D during the transfer from pediatric to adult care. MethodsOne-hundred young adults (age 17–25) with T1D and 29 Peer Mentors enrolled in this randomized clinical trial. Peer Mentors are experienced, older young adults with T1D, trained by the study team to share transition experiences and strategies to successfully navigate the adult healthcare system, help young adults prepare for the first adult care visit, and use strengths-based support strategies to teach and model skills for managing T1D-related challenges. ResultsThe primary outcome of the trial is HbA1c, and secondary outcomes include time to adult care, engagement in diabetes self-management behaviors, and psychosocial well-being. ConclusionThe goal of this research is to evaluate a developmentally appropriate, supportive intervention that can improve T1D self-management and successful transfer of care during the difficult young adult years and promote optimal T1D health outcomes.

Ethnicity Detection with Convolutional Neural Network (CNN): Bangladesh Perspectives

Ethnicity detection, or the automatic determination of people's ethnic backgrounds using visual data, has become important in a number of disciplines, including social sciences, demographics, and computer vision. With a focus on Bangladesh's complex cultural landscape, we describe in this paper a novel method for ethnicity detection using convolutional neural networks (CNNs). We gathered a dataset of photographs depicting members of the three main ethnic groups living in the Chittagong hill neighborhood: Chakma, Marma, and Tripura. Utilizing the DenseNet121 model's robust feature extraction skills, we customized the architecture to meet the needs of our particular ethnicity detection application. A wide range of performance indicators were used to train and assess our unique CNN model. The outcomes illustrate accurate ethnicity detection technology's potential for identifying and resolving social inequities while showcasing a promising level of accuracy. To address possible biases in the model's prediction, ethical considerations are also covered. Overall, this work advances knowledge of Bangladesh's ethnic variety and demonstrates the potential of CNNs for ethnicity detection. The results provide up new avenues for investigation and applications that advance a more diverse and inclusive society.

Developing modelling competencies in pre-service mathematics teachers through reflective thinking skills

Background: Ghana’s commitment to quality education has been reflected in its goal of providing equal access to high-quality education, leading to the reforms of the New Common Core Mathematics Curriculum for Basic Schools. The study explored the integration of mathematical modelling using reflective thinking skills, which are not currently core competencies in Ghanaian basic school curriculum.Aim: The study examined the modelling proficiency of pre-service mathematics teachers by assessing their reflective thinking abilities in a modelling laboratory context.Setting: The study focussed on pre-service mathematics teachers from two education colleges in Ghana.Methods: Using purposive sampling to select participants, a quasi-experimental design with pre- and post-test interventions was employed. Data were analysed through content and inferential analysis, supplemented by interviews.Findings: The findings indicated that the comparison group lacked prior knowledge of modelling problems and struggled with comprehension tasks. In contrast, the experimental group successfully translated real-world problems into mathematical models.Conclusion: Providing pre-service mathematics teachers access to a modelling laboratory and modelling-eliciting activities was essential for developing future modellers. This approach would enhance their effectiveness in teaching foundational mathematics in Ghanaian education.Contribution: This study advocated for re-orienting the mathematics curriculum at both Basic schools and Colleges of Education in Ghana to include mathematical modelling and reflective thinking skills as core components.

A phenotypic drug discovery approach by latent interaction in deep learning

Contemporary drug discovery paradigms rely heavily on binding assays about the bio-physicochemical processes. However, this dominant approach suffers from overlooked higher-order interactions arising from the intricacies of molecular mechanisms, such as those involving cis -regulatory elements. It introduces potential impairments and restrains the potential development of computational methods. To address this limitation, I developed a deep learning model that leverages an end-to-end approach, relying exclusively on therapeutic information about drugs. By transforming textual representations of drug and virus genetic information into high-dimensional latent representations, this method evades the challenges arising from insufficient information about binding specificities. Its strengths lie in its ability to implicitly consider complexities such as epistasis and chemical–genetic interactions, and to handle the pervasive challenge of data scarcity. Through various modeling skills and data augmentation techniques, the proposed model demonstrates outstanding performance in out-of-sample validations, even in scenarios with unknown complex interactions. Furthermore, the study highlights the importance of chemical diversity for model training. While the method showcases the feasibility of deep learning in data-scarce scenarios, it reveals a promising alternative for drug discovery in situations where knowledge of underlying mechanisms is limited.