QSAR: Using the Past to Study the Present.

Quantitative structure-activity relationships (QSAR) is a method for predicting the physical and biological properties of small molecules; it is in use in industry and public services. However, as any scientific method, it is challenged by more and more requests, especially considering its possible role in assessing the safety of new chemicals. To answer the question whether QSAR, by exploiting available knowledge, can build new knowledge, the chapter reviews QSAR methods in search of a QSAR epistemology. QSAR stands on tree pillars, i.e., biological data, chemical knowledge, and modeling algorithms. Usually the biological data, resulting from good experimental practice, are taken as a true picture of the world; chemical knowledge has scientific bases; so if a QSAR model is not working, blame modeling. The role of modeling in developing scientific theories, and in producing knowledge, is so analyzed. QSAR is a mature technology and is part of a large body of in silico methods and other computational methods. The active debate about the acceptability of the QSAR models, about the way to communicate them, and the explanation to provide accompanies the development of today QSAR models. An example about predicting possible endocrine-disrupting chemicals (EDC) shows the many faces of modern QSAR methods.

Machine Learning Elucidates Design Features of Plasmid Deoxyribonucleic Acid Lipid Nanoparticles for Cell Type-Preferential Transfection.

To broaden the accessibility of cell and gene therapies, it is essential to develop and optimize nonviral, cell type-preferential gene carriers such as lipid nanoparticles (LNPs). While high-throughput screening (HTS) approaches have proven effective in accelerating LNP discovery, they are often costly, labor-intensive, and do not consistently yield actionable design rules that direct screening efforts toward the most relevant chemical and formulation parameters. In this study, we employed a machine learning (ML) workflow, utilizing well-curated plasmid DNA LNP transfection data sets across six cell types, to extract compositional and chemical insights from HTS studies. Our approach achieved prediction errors averaging between 5 and 10%, depending on the cell type. By applying SHapley Additive exPlanations to our ML models, we uncovered key composition-function relationships that govern cell type-preferential LNP transfection efficiency. Notably, we identified consistent LNP composition parameters that enhance in vitro transfection efficiency across diverse cell types, including a helper lipid molar percentage of charged lipids between 9 and 50% and the inclusion of cationic/zwitterionic helper lipids. Additionally, several parameters were found to modulate cell type-preferentiality, such as the total molar percentage of ionizable and helper lipids, N/P ratio, PEGylated lipid molar percentage of uncharged lipids, and hydrophobicity of the helper lipid. This study leverages HTS of compositionally diverse LNP libraries combined with ML analysis to elucidate the interactions between lipid components in LNP formulations, providing insights that contribute to the design of LNP compositions tailored for cell type-preferential transfection.

Doing things efficiently: Testing an account of why simple explanations are satisfying.

People often find simple explanations more satisfying than complex ones. Across seven preregistered experiments, we provide evidence that this simplicity preference is not specific to explanations and may instead arises from a broader tendency to prefer completing goals in efficient ways. In each experiment, participants (total N=2820) learned of simple and complex methods for producing an outcome, and judged which was more appealing-either as an explanation why the outcome happened, or as a process for producing it. Participants showed similar preferences across judgments. They preferred simple methods as explanations and processes in tasks with no statistical information about the reliability or pervasiveness of causal elements. But when this statistical information was provided, preferences for simple causes often diminished and reversed in both kinds of judgments. Together, these findings suggest that people may assess explanations much in the same ways they assess methods for completing goals, and that both kinds of judgments depend on the same cognitive mechanisms.

Evaluating machine learning performance in predicting sodium adsorption ratio for sustainable soil-water management in the eastern Mediterranean.

Soil salinization is a critical global issue for sustainable agriculture, impacting crop yields and posing a threat to achieving the Sustainable Development Goal (SDG) of ensuring food security. It is necessary to monitor it in detail and uncover its underlying factors at a regional scale. In this context, the present study aimed to evaluate soil health in the eastern Mediterranean region by using the Sodium Adsorption Ratio (SAR) as an indicator of soil salinity in three distinct soil horizons. The main objective of the research was to evaluate the performance of four machine learning (ML) models, including Random Forest (RF), Nu Support Vector Regression (NuSVR), Artificial Neural Network-Multi Layer Perceptron (ANN-MLP), and Gradient Boosting Regression (GBR), for accurate prediction of SAR following the Recursive Feature Elimination (RFE) as a feature selection method. Moreover, SHapely Additive exPlanations (SHAP) was applied as sensitivity analysis to identify the most influential covariates. Main findings of the research revealed that the average clay content in the surface horizon (H10-25cm) was 50.5% ± 10.4, which significantly increased to 57.5% ± 8.7 (p < 0.05). No significant mean differences were detected between the studied horizons for SAR and Na+. ML output revealed that NuSVR outperformed other algorithms in accurately predicting outcomes during both the training and testing stages. Moreover, Scenario 2 (SC2) with seven selected features from the RFE method facilitated highly accurate SAR predictions. Overall, the performance of ML models is ranked as NuSVR > GBR > ANN-MLP > RF. Lastly, SHAP sensitivity analysis identified CEC, Ca+2, Mg+2, and Na+ as the most influential variables for SAR prediction in both the training and testing stages. Hence, the research yielded valuable insights for efficient agricultural soil management at a regional level using state-of-the-art technology.

Grounding, necessity, and relevance.

Grounding necessitarianism (GN) is the view that full grounds necessitate what they ground. Although GN has been rather popular among philosophers, it faces important counterexamples: For instance, A = [Socrates died] fully grounds C = [Xanthippe became a widow]. However, A fails to necessitate C: A could have obtained together with B = [Socrates and Xanthippe were never married], without C obtaining. In many cases, the debate essentially reduces to whether A indeed fully grounds C-as the contingentist claims-or if instead C is fully grounded in A+, namely A plus some supplementary fact S (e.g. [Xanthippe was married to Socrates])-as the necessitarian claims. Both sides typically agree that A+ necessitates C, while A does not; they disagree on whether A or A+ fully grounds C. This paper offers a novel defence of the claim that, in these typical cases, unlike A+, A fails to fully ground C-thereby bringing further support to GN. First and foremost, unlike A+, A fails to fully ground C because it fails to contain just what is relevant to do so, in two distinct senses-explanatory and generative relevance. Second, going for A, rather than A+, as a full ground undermines not just grounding necessitarianism, but modally weaker views which even contingentists may want to preserve.

Interpretable machine learning model predicting immune checkpoint inhibitor-induced hypothyroidism: A retrospective cohort study.

Hypothyroidism is a known adverse event associated with the use of immune checkpoint inhibitors (ICIs) in cancer treatment. This study aimed to develop an interpretable machine learning (ML) model for individualized prediction of hypothyroidism in patients treated with ICIs. The retrospective cohort of patients treated with ICIs was from the First Affiliated Hospital of Ningbo University. ML methods applied include logistic regression (LR), random forest classifier (RFC), support vector machine (SVM), and extreme gradient boosting (XGBoost). The area under the receiver-operating characteristic curve (AUC) was the main evaluation metric used. Furthermore, the Shapley additive explanation (SHAP) was utilized to interpret the outcomes of the prediction model. A total of 458 patients were included in the study, with 59 patients (12.88%) observed to have developed hypothyroidism. Among the models utilized, XGBoost exhibited the highest predictive capability (AUC = 0.833). The Delong test and calibration curve indicated that XGBoost significantly outperformed the other models in prediction. The SHAP method revealed that thyroid-stimulating hormone (TSH) was the most influential predictor variable. The developed interpretable ML model holds potential for predicting the likelihood of hypothyroidism following ICI treatment in patients. ML technology offers new possibilities for predicting ICI-induced hypothyroidism, potentially providing more precise support for personalized treatment and risk management.

Understanding, fast and shallow: Individual differences in memory performance associated with cognitive load predict the illusion of explanatory depth.

People are often overconfident about their ability to explain how everyday phenomena and artifacts work (devices, natural processes, historical events, etc.). However, the metacognitive mechanisms involved in this bias have not been fully elucidated. The aim of this study was to establish whether the ability to perform deliberate and analytic processes moderates the effect of informational cues such as the social desirability of knowledge on the Illusion of Explanatory Depth (IOED). To this purpose, the participants' cognitive load was manipulated as they provided initial estimates of causal understanding of national historical events in the standard IOED paradigm. The results showed that neither the social desirability of specific causal knowledge nor the cognitive load manipulations had direct effects on the IOED. However, subsequent exploratory analyses indicated that high cognitive load was related to lower performance on concurrent memory tasks, which in turn was associated with a higher IOED magnitude. Higher analytical processing was also related to lower IOED. Implications for both dual-process models of metacognition and the design of task environments that help to reduce this bias are discussed.

A study of dentists' explanations and patient-dentist communication among older adults in Japan.

BACKGROUND: In recent years, the proportion of elderly people in the total patient population has been increasing owing to the rapid aging of Japanese society. However, little is known about the age-specific healthcare communication challenges within the field of dentistry. Therefore, this study aimed to examine the relationship between dentists' explanations and patient-dentist communication among elderly patients. METHODS: The study included 146 dentist-elderly patient pairs from Fukuoka Prefecture, Japan. A questionnaire was administered to pairs of dentists and patients. The survey was conducted between June 2021 and April 2022. We examined the relationships among the survey items: dentist demographics, patient demographics and sufficiency of the dentist explanations, and patient-dentist communication. The logistic regression analysis was conducted to examine the patient's mode of visiting the dentist as the objective variable, sufficiency of the dentist explanations, patient-dentist communication, dentist, and patient factors as explanatory variables. RESULTS: About 30% of patients felt that explanations of "Comparison with other treatment methods," "Treatment period," and "Treatment prognosis" were not sufficient. Among these items, a significantly higher percentage of respondents found the dentist's explanations sufficient when they were treated by more than one dentist. Many good communication factors were significantly associated with the dentist being younger and having a preventive practice. Multivariate analysis revealed a significant association between sufficiency of the dentist explanations and patients' regular dental visits. CONCLUSION: Adequate explanations by dentists for elderly patients were significantly associated with the dentist factor. Improving the quantity and quality of the dentists' explanations of treatment may improve patient satisfaction and promote regular dental visits.

Embodied (4EA) cognitive computational neuroscience.

I argue that ideas and models about the mechanisms of neural computation and representation - including computational architecture, representational format, encoding schemes, learning methods, computation-representation coordination, and substrate-dependent aspects - must be tested by studying embodied neural systems. Thus, cognitive computational neuroscience - the study of neural computations over neural representations - must be an embodied research program.

Beyond embodiment: Rethinking the integration of cognitive neuroscience and mechanistic explanations.

This commentary critiques Mougenot and Matheson's proposal to integrate embodied cognition with mechanistic explanations in cognitive neuroscience. We suggest more promising directions for embodied cognitive neuroscience, focusing on neuroethological research and evolutionary studies of nervous systems. These approaches, compatible with wide mechanistic explanations, offer a robust path forward by examining central nervous system function within whole organisms in their environments.

Learning by thinking in natural and artificial minds.

Canonical cases of learning involve novel observations external to the mind, but learning can also occur through mental processes such as explaining to oneself, mental simulation, analogical comparison, and reasoning. Recent advances in artificial intelligence (AI) reveal that such learning is not restricted to human minds: artificial minds can also self-correct and arrive at new conclusions by engaging in processes of 'learning by thinking' (LbT). How can elements already in the mind generate new knowledge? This article aims to resolve this paradox, and in so doing highlights an important feature of natural and artificial minds - to navigate uncertain environments with variable goals, minds with limited resources must construct knowledge representations 'on demand'. LbT supports this construction.

"Social science is explanation or it is nothing." Introduction to a debate.

This essay introduces contributions to a special section, which documents and extends a debate on the proposition "Social Science is Explanation or it is Nothing" held at the London School of Economics on October 13th, 2022. It discusses the history of the "Group for Theoretical Debates in Anthropology" led by Tim Ingold, Peter Wade and Soumhya Venkatesan, which has handed down a list of credible candidates for issues that had a chance of engaging every anthropologist, including students and those with interdisciplinary interests. It raises questions about the specific affordances of debates as forms of academic engagements. It argues that the chosen proposition concerning explanation invites a discussion about the contributions of the social sciences at a time when impulses from science and technology studies as well as fruitful exchanges across the boundary between "theory" and "method" have helped us moved beyond the older question as to whether or not sociology is "a science".

The Application of Machine Learning Algorithms to Bond Strength between Steel Rebars and Concrete Using Bayesian Optimization.

The purpose of this study is to estimate the bond strength between steel rebars and concrete using machine learning (ML) algorithms with Bayesian optimization (BO). It is important to conduct beam tests to determine the bond strength since it is affected by stress fields. A machine learning approach for bond strength based on 401 beam tests with six impact factors is presented in this paper. The model is composed of three standard algorithms, including random forest (RF), support vector regression (SVR), and extreme gradient boosting (XGBoost), combined with the BO technique. Compared to empirical models, BO-XGB`oost was found to be the most accurate method, with values of R2, MAE, and RMSE of 0.87, 0.897 MPa, and 1.516 MPa for the test set. The development of a simplified model that contains three input variables (diameter of the rebar, yield strength of reinforcement, concrete compressive strength) has been proposed to make it more convenient to apply. According to this prediction, the Shapley additive explanation (SHAP) can help explain why the ML-based model predicts the particular outcome it does. By utilizing machine learning algorithms to predict complex interfacial mechanical behavior, it is possible to improve the accuracy of the model.

What is social science if not critical?

This short article represents a contribution to the debate on the motion "Social science is explanation, or it is nothing." While in the format of parliamentary debating the contribution would fall on the side of the opposition, I will not be arguing against explanation as such. The work of explaining is in no way oppositional to or mutually exclusive with critique. Instead, my contribution will revolve around two arguments: one is that both critique and explanation exhibit characteristics we commonly attribute to science; the other is that reserving the label of science for explanation draws a boundary around social sciences in ways that exclude many of the interesting things it does. Some of the examples include the sociological analysis of governmental approaches to the COVID-19 pandemic, or critical analysis of concepts such as "cancel culture" or "terrorism." The conclusion is that explanation and critique are mutually supporting elements of science, and that combined they give us insights we cannot glean from either alone.

An early sepsis prediction model utilizing machine learning and unbalanced data processing in a clinical context.

Background: Early and accurate diagnoses of sepsis patients are essential to reduce the mortality. However, the sepsis is still diagnosed in a traditional way in China despite the increasing number of related studies, which may to some extent lead to delays in the treatment. Methods: The study included 2,385 patients, including 364 with sepsis, collected from the First Affiliated Hospital of Anhui Medical University and partner hospitals from April to July 2022. External validation was conducted using the MIMIC-III database (over 60,000 patients from 2001 to 2012) and the eICU Collaborative Research Database (139,000 patients from 2014 to 2015). Multiple algorithm models, along with the SHapley Additive exPlanations (SHAP) analysis, are applied to explore the main risk factors for the accurate prediction of the sepsis. Multiple Imputations for filling missing data and the Synthetic Minority Oversampling (SMOTE) balancing method for balancing data are used for the data processing. Result: Eighteen diagnostic features are used in the predictive model for early sepsis. The Random Forest model has the best performance among all the models, with an Area Under the Curve (AUC) of 87% and an F1-score (F1) of 77%. Moreover, the interpretation from the SHAP analysis is generally consistent with the current clinical situation. Conclusion: The study revealed the relationship between these 18 clinical features and diagnostic outcomes. The results indicate that patients with laboratory values of Systolic Blood Pressure, Albumin, and Heart Rate exceeding certain thresholds are at a high likelihood of developing sepsis.

Manifold-based Shapley explanations for high dimensional correlated features.

Explainable artificial intelligence (XAI) holds significant importance in enhancing the reliability and transparency of network decision-making. SHapley Additive exPlanations (SHAP) is a game-theoretic approach for network interpretation, attributing confidence to inputs features to measure their importance. However, SHAP often relies on a flawed assumption that the model's features are independent, leading to incorrect results when dealing with correlated features. In this paper, we introduce a novel manifold-based Shapley explanation method, termed Latent SHAP. Latent SHAP transforms high-dimensional data into low-dimensional manifolds to capture correlations among features. We compute Shapley values on the data manifold and devise three distinct gradient-based mapping methods to transfer them back to the high-dimensional space. Our primary objectives include: (1) correcting misinterpretations by SHAP in certain samples; (2) addressing the challenge of feature correlations in high-dimensional data interpretation; and (3) reducing algorithmic complexity through Manifold SHAP for application in complex network interpretations. Code is available at https://github.com/Teriri1999/Latent-SHAP.

Evidence modeling for reliability learning and interpretable decision-making under multi-modality medical image segmentation.

Reliability learning and interpretable decision-making are crucial for multi-modality medical image segmentation. Although many works have attempted multi-modality medical image segmentation, they rarely explore how much reliability is provided by each modality for segmentation. Moreover, the existing approach of decision-making such as the softmax function lacks the interpretability for multi-modality fusion. In this study, we proposed a novel approach named contextual discounted evidential network (CDE-Net) for reliability learning and interpretable decision-making under multi-modality medical image segmentation. Specifically, the CDE-Net first models the semantic evidence by uncertainty measurement using the proposed evidential decision-making module. Then, it leverages the contextual discounted fusion layer to learn the reliability provided by each modality. Finally, a multi-level loss function is deployed for the optimization of evidence modeling and reliability learning. Moreover, this study elaborates on the framework interpretability by discussing the consistency between pixel attribution maps and the learned reliability coefficients. Extensive experiments are conducted on both multi-modality brain and liver datasets. The CDE-Net gains high performance with an average Dice score of 0.914 for brain tumor segmentation and 0.913 for liver tumor segmentation, which proves CDE-Net has great potential to facilitate the interpretation of artificial intelligence-based multi-modality medical image fusion.

Spatio-temporal learning and explaining for dynamic functional connectivity analysis: Application to depression.

BACKGROUND: Functional connectivity has been shown to fluctuate over time. The present study aimed to identifying major depressive disorders (MDD) with dynamic functional connectivity (dFC) from resting-state fMRI data, which would be helpful to produce tools of early depression diagnosis and enhance our understanding of depressive etiology. METHODS: The resting-state fMRI data of 178 subjects were collected, including 89 MDD and 89 healthy controls. We propose a spatio-temporal learning and explaining framework for dFC analysis. A yet effective spatio-temporal model is developed to classifying MDD from healthy controls with dFCs. The model is a stacking neural network model, which learns network structure information by a multi-layer perceptron based spatial encoder, and learns time-varying patterns by a Transformer based temporal encoder. We propose to explain the spatio-temporal model with a two-stage explanation method of importance feature extracting and disorder-relevant pattern exploring. The layer-wise relevance propagation (LRP) method is introduced to extract the most relevant input features in the model, and the attention mechanism with LRP is applied to extract the important time steps of dFCs. The disorder-relevant functional connections, brain regions, and brain states in the model are further explored and identified. RESULTS: We achieved the best classification performance in identifying MDD from healthy controls with dFC data. The top important functional connectivity, brain regions, and dynamic states closely related to MDD have been identified. LIMITATIONS: The data preprocessing may affect the classification performance of the model, and this study needs further validation in a larger patient population. CONCLUSIONS: The experimental results demonstrate that the proposed spatio-temporal model could effectively classify MDD, and uncover structural and temporal patterns of dFCs in depression.

Exploring high-emission driving behaviors of heavy-duty diesel vehicles based on engine principles under different road grade levels.

To reveal the outstanding high-emission problems that occur when heavy-duty diesel vehicles (HDDV) pass uphill and downhill, this study proposes a method to depict the nitrogen oxides (NOx) and carbon dioxide (CO2) high-emission driving behaviors caused by slopes from the perspective of engine principles. By calculating emission and grade data of HDDV based on on-board diagnostic (OBD) data and digital elevation model (DEM) data, the 262 short trips including uphill, flat-road and downhill are firstly obtained through the rule-based short trip segmentation method, and the significant correlation between the road grade and emissions of the short trips is verified by Kendall's Tau and K-means clustering. Secondly, by comparing the distribution changes of three speed categories (acceleration state, constant speed state and deceleration state), the differences in HDDV operating states under different grade levels are discussed. Finally, the machine learning models (Random Forest, XGBoost and Elastic Net), are used to develop the NOx and CO2 emission estimation model, identifying high-emission driving behaviors, particularly during uphill driving, which showed the highest proportion of high-emission. Explained by the feature importance and SHapley Additive exPlanations (SHAP) model that large accelerator pedal opening, frequent aggressive acceleration, and high engine load have positive effects both on NOx and CO2 emissions. The difference is in the air-fuel ratio that the engine in the rich or slightly lean burning state will increase CO2 emissions and the lean burning state will increase NOx emissions. In addition, due to the uncertainty of the actual uphill, drivers often undergo a rapid "deceleration-uniform-acceleration" process, which significantly contributes to high NOx and CO2 emissions from the engine perspective. The findings provide insights for designing driving strategies in slope scenarios and offer a novel perspective on depicting driving behaviors.

Inference to the best neuroscientific explanation.

Neuroscientists routinely use reverse inference (RI) to draw conclusions about cognitive processes from neural activation data. However, despite its widespread use, the methodological status of RI is a matter of ongoing controversy, with some critics arguing that it should be rejected wholesale on the grounds that it instantiates a deductively invalid argument form. In response to these critiques, some have proposed to conceive of RI as a form of abduction or inference to the best explanation (IBE). We side with this response but at the same time argue that a defense of RI requires more than identifying it as a form of IBE. In this paper, we give an analysis of what determines the quality of an RI conceived as an IBE and on that basis argue that whether an RI is warranted needs to be decided on a case-by-case basis. Support for our argument will come from a detailed methodological discussion of RI in cognitive neuroscience in light of what the recent literature on IBE has identified as the main quality indicators for IBEs.