Accelerating Biosensor Discovery: A Computationally‐Driven Pipeline for Microplastics Monitoring

Important structural parameter for curvature effect of TM-N4 embeded C70 fullerenes as electrocatalysts for CO2 reduction interpreted with machine learning and first-principles calculations

Implementing and understanding the unsupervised transfer learning in metal organic framework toward methane adsorption from hypothetical to experimental data

The virtual screening (VS) is an important tool used in the modern drug discovery process to identify new leads and drug-like molecules for therapeutic interventions. With the rapid advancements in the computational hardware supported by advanced algorithmic progress and comprehensive data, the VS has gained fast momentum in drug discovery paradigm. The VS protocol is performed predominantly by using the ligand-based (LB) and structure-based (SB) VS methods with each one having its own merits and demerits. The LBVS method works on the similarity approach based on the physicochemical parameters, chemical functionality, and shape similarity of the ligands. The SBVS method works on the complementarity of the ligand with the target binding site and is more preferred than LBVS due to the consideration of both ligand and target information. Further in the SBVS, the available conformational information of the active chemical space helps in making reasonable decisions for the ligand selection. However, the limitations in scoring functions and incorrect pose prediction may result in the inaccurate SB models of limited performance in VS experiments. Various factors interplay in the development of successful VS models and the fine tuning of these factors leads to the efficient VS models with high sensitivity. In this chapter, some of the recent ligand- and structure-based approaches for virtual screening in drug discovery are described.

The present study aimed to develop robust machine learning (ML) models to predict the skin permeability of poorly water-soluble drugs in the presence of menthol and limonene as penetration enhancers (PEs). The ML models were also applied in virtual screening (VS) to identify hydrophobic drugs that exhibited better skin permeability in the presence of permeation enhancers i.e. menthol and limonene. The drugs identified through ML-based VS underwent experimental validation using in vitro skin penetration studies. The developed model predicted 80% probability of permeability enhancement for Sumatriptan Succinate (SS), Voriconazole (VCZ), and Pantoprazole Sodium (PS) with menthol and limonene. The in vitro release studies revealed that menthol increased penetration by approximately 2.49-fold, 2.25-fold, and 4.96-fold for SS, VCZ, and PS, respectively, while limonene enhanced permeability by approximately 1.32-fold, 2.27-fold, and 3.7-fold for SS, VCZ, and PS. The results from in silico and in vitro studies were positively correlated, indicating that the developed ML models could effectively reduce the need for extensive in vitro and in vivo experimentation.

Electroreduction of nitrate to ammonia on graphyne-based single-atom catalysts by combined density functional theory and machine learning study

It has been demonstrated that many promising thermoelectric materials, such as tetradymite compounds are also three-dimensional topological insulators. In both cases, a fundamental question is the evaluation of carrier relaxation time, which is usually a rough task due to the complicated scattering mechanisms. Previous works using the simple deformation potential theory or considering complete electron-phonon coupling are, however, restricted to small systems. By adopting a data-driven method named SISSO (Sure Independence Screening and Sparsifying Operator) with the training data obtained via deformation potential theory, we propose an efficient and physically interpretable descriptor to evaluate the relaxation time, using tetradymites as prototypical examples. Without any input from first-principles calculations, the descriptor contains only several elemental properties of the constituent atoms, and could be utilized to quickly and reliably predict the carrier relaxation time of a substantial number of tetradymites with arbitrary stoichiometry.

Data-driven engineering descriptor and refined scale relations for predicting bubble departure diameter

Perovskite-type lead-free piezoelectric ceramics allow access to illustrious piezoelectric coefficients (d33) through intricate composition design and experimental modulation. Developing a swift and accurate technology for identifying (K, Na)NbO3 (KNN)-based ceramic compositions with high d33 in exceedingly large “compositional” space will establish an innovative research paradigm surpassing the traditional empirical trial-and-error method. Herein, we demonstrate an interpretable machine learning (ML) framework for quick evaluation of KNN-based ceramics with high d33 based on data from published literature. Specifically, a thorough feature construction was carried out from the global and local dimensions to establish tree regression models with d33 as the target property. Subsequently, the feature-property mapping rules of KNN-based piezoelectric ceramics are further optimized through feature screening. To intuitively understand the correlation mechanisms between ML regression targets and features, the sure independence screening and sparsifying operator (SISSO) method was employed to extract the essential descriptors to explain d33. A straightforward descriptor, e(NMB−MVB)⋅ST/(IDA)2\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$${\ ext{e}}^{({{NM}}_{\ ext{B}}-{{MV}}_{\ ext{B}})}\\cdot {ST}/{(I{D}_{\ ext{A}})}^{2}$$\\end{document}, consisting of only four easily accessible parameters, can accelerate the evaluation of a series of novel KNN-based ceramics with high d33 while exhibiting strong theoretical interpretability. This work not only provides a tool for the rapid discovery of high piezoelectric performance in KNN-based ceramics but also offers a data-driven route for the design of property descriptors in perovskites.

The field of materials science has experienced a transformative shift with the emergence of high-entropy materials (HEMs), which possess a unique combination of properties that traditional single-phase materials lack. Among these, high-entropy nitrides (HENs) stand out for their exceptional mechanical strength, thermal stability, and resistance to extreme environments, making them highly sought after for applications in aerospace, defense, and energy sectors. Central to the design of these materials is their entropy forming ability (EFA), a measure of a material’s propensity to form a single-phase, disordered structure. This study introduces the application of the sure independence screening and sparsifying operator (SISSO), a machine learning technique, to predict the EFA of HEN ceramics. By utilizing a rich dataset curated from theoretical computational data, SISSO has been trained to identify the most critical features contributing to EFA. The model’s strong interpretability allows for the extraction of complex mathematical expressions, providing deep insights into the material’s composition and its impact on EFA. The predictive performance of the SISSO model is meticulously validated against theoretical benchmarks and compared with other machine learning methodologies, demonstrating its superior accuracy and reliability. This research not only contributes to the growing body of knowledge on HEMs but also paves the way for the efficient discovery and development of new HEN materials with tailored properties for advanced technological applications.

Ulcerative colitis (UC) is a chronic gastrointestinal inflammatory disorder with rising prevalence. Due to the recurrent and difficult-to-treat nature of UC symptoms, current pharmacological treatments fail to meet patients' expectations. This study presents a machine learning-assisted high-throughput screening strategy to expedite the discovery of efficient nanozymes for UC treatment. Therapeutic requirements, including antioxidant property, acid stability, and zeta potential, are quantified and predicted by using a machine learning model. Non-quantifiable attributes, including intestinal barrier repair efficacy and biosafety, are assessed via high-throughput screening. Feature significance analysis, sure independence screening, and sparsifying operator symbolic regression reveal the high-dimensional structure-activity relationships between material features and therapeutic needs. SrDy2O4 with high stability, low toxicity, targeting ability, and reactive oxygen species (ROS) scavenging capability is identified, which reduces ROS production, lowers cytochrome C levels in cytoplasm, and inhibits apoptosis in intestinal epithelial cells by stabilizing the mitochondrial membrane potential. Mice treated with SrDy2O4 show improvements in colon length and body weight compared with dextran sodium sulfate salt-treated model group. Transcriptomic and 16S rRNA sequencing analyses show that SrDy2O4 boosts beneficial gut bacteria, and decreases pathogenic bacteria, thereby effectively restoring gut microbiota balance. Moreover, SrDy2O4 offers the advantage of X-ray imaging without side effects.

N-heterocyclic carbenes (NHCs) are very versatile ligands known for their strong σ-donating ability and stability, making them suitable for forming complexes with alkaline earth metals (AEMs). In this study, we designed a machine learning (ML) based approach for high-throughput screening of NHC-AEM complexes based on their binding energies (BEs), obtained from density functional theory (DFT) optimized structures and single-point domain-based local pair-natural orbital based singles and doubles coupled cluster (DLPNO-CCSD(T)) calculations. The ML models were trained on descriptors derived from fundamental metal properties, RDKit and Sterimol parameters, which were refined through recursive feature refinement steps. The ML workflow involved successive feature refinement steps, beginning with principle component analysis (PCA) filtering (r > 0.9), feature engineering (only in the case of Sterimol parameters), sure independence screening (SIS) based on distance correlation, and stepwise feature elimination. Seven different ML models were trained on the different datasets mentioned above, with the best one achieving a mean square error (MSE) of 0.10 eV2. Furthermore, Shapley additive explanations (SHAP) analysis was employed to identify the most influential descriptor governing BE.

Small data, big challenges: Machine- and deep-learning strategies for data-limited drug discovery.

Virtual Screening (VS) methods aid clinical research by predicting the interaction of ligands with pharmacological targets, thus accelerating the process of finding new drugs. The computational requirements of VS models, along with the size of the databases, containing up to millions of biological macromolecular structures, propitiates the use of High-Performance Computing. METADOCK is a tool for the application of metaheuristics to VS in heterogeneous clusters of computers based on CPU and GPU. It facilitates the efficient application of several metaheuristics in parallel computational systems. HYPERDOCK represents a step forward; the exploration for satisfactory metaheuristics is systematically approached by means of hyperheuristics working on top of the metaheuristic schema of METADOCK. The parallelism of METADOCK is intrinsically exploited in HYPERDOCK, which also includes parallelism at its own level. HYPERDOCK helps to generate better ligand-receptor bindings.

Quantitative Structure Activity Relationship (QSAR) models can inform on the correlation between activities and structure-based molecular descriptors. This information is important for the understanding of the factors that govern molecular properties and for designing new compounds with favorable properties. Due to the large number of calculate-able descriptors and consequently, the much larger number of descriptors combinations, the derivation of QSAR models could be treated as an optimization problem. For continuous responses, metrics which are typically being optimized in this process are related to model performances on the training set, for example, and . Similar metrics, calculated on an external set of data (e.g., ), are used to evaluate the performances of the final models. A common theme of these metrics is that they are context -” ignorant”. In this work we propose that QSAR models should be evaluated based on their intended usage. More specifically, we argue that QSAR models developed for Virtual Screening (VS) should be derived and evaluated using a virtual screening-aware metric, e.g., an enrichment-based metric. To demonstrate this point, we have developed 21 Multiple Linear Regression (MLR) models for seven targets (three models per target), evaluated them first on validation sets and subsequently tested their performances on two additional test sets constructed to mimic small-scale virtual screening campaigns. As expected, we found no correlation between model performances evaluated by “classical” metrics, e.g., and and the number of active compounds picked by the models from within a pool of random compounds. In particular, in some cases models with favorable and/or values were unable to pick a single active compound from within the pool whereas in other cases, models with poor and/or values performed well in the context of virtual screening. We also found no significant correlation between the number of active compounds correctly identified by the models in the training, validation and test sets. Next, we have developed a new algorithm for the derivation of MLR models by optimizing an enrichment-based metric and tested its performances on the same datasets. We found that the best models derived in this manner showed, in most cases, much more consistent results across the training, validation and test sets and outperformed the corresponding MLR models in most virtual screening tests. Finally, we demonstrated that when tested as binary classifiers, models derived for the same targets by the new algorithm outperformed Random Forest (RF) and Support Vector Machine (SVM)-based models across training/validation/test sets, in most cases. We attribute the better performances of the Enrichment Optimizer Algorithm (EOA) models in VS to better handling of inactive random compounds. Optimizing an enrichment-based metric is therefore a promising strategy for the derivation of QSAR models for classification and virtual screening.

Autism Spectrum Disorder (ASD) is a lifelong developmental condition for which early identification is essential. The Autism Quotient-10 (AQ-10) is frequently used as a quick screening instrument, classifying individuals scoring seven or more as potentially autistic. This study applies machine learning (ML) models to a public AQ-10 dataset to determine whether predictive algorithms can provide insights beyond the rule-based threshold. Logistic Regression and Random Forest classifiers were built, validated, and compared. Both models demonstrated perfect scores on all evaluation metrics, reflecting the deterministic nature of the dataset rather than offering new diagnostic information. Demographic examination revealed meaningful differences in screening patterns across gender, age groups, ethnicity, and family history of ASD. These findings emphasize that ML models trained on threshold-derived labels cannot infer clinical patterns beyond the embedded scoring rule. Future work will require clinician-confirmed datasets and multimodal features to meaningfully advance computational autism screening.