A scoping review of guidelines for the use of race, ethnicity, and ancestry reveals widespread consensus but also points of ongoing disagreement.

The use of population descriptors such as race, ethnicity, and ancestry in science, medicine, and public health has a long, complicated, and at times dark history, particularly for genetics, given the field's perceived importance for understanding between-group differences. The historical and potential harms that come with irresponsible use of these categories suggests a clear need for definitive guidance about when and how they can be used appropriately. However, while many prior authors have provided such guidance, no established consensus exists, and the extant literature has not been examined for implied consensus and sources of disagreement. Here, we present the results of a scoping review of published normative recommendations regarding the use of population categories, particularly in genetics research. Following PRISMA guidelines, we extracted recommendations from n = 121 articles matching inclusion criteria. Articles were published consistently throughout the time period examined and in a broad range of journals, demonstrating an ongoing and interdisciplinary perceived need for guidance. Examined recommendations fall under one of eight themes identified during analysis. Seven are characterized by broad agreement across articles; one, "appropriate definitions of population categories and contexts for use," revealed substantial fundamental disagreement among articles. Additionally, while many articles focus on the inappropriate use of race, none fundamentally problematize ancestry. This work can be a resource to researchers looking for normative guidance on the use of population descriptors and can orient authors of future guidelines to this complex field, thereby contributing to the development of more effective future guidelines for genetics research.

Fusing 2D and 3D molecular graphs as unambiguous molecular descriptors for conformational and chiral stereoisomers.

The rapid progress of machine learning (ML) in predicting molecular properties enables high-precision predictions being routinely achieved. However, many ML models, such as conventional molecular graph, cannot differentiate stereoisomers of certain types, particularly conformational and chiral ones that share the same bonding connectivity but differ in spatial arrangement. Here, we designed a hybrid molecular graph network, Chemical Feature Fusion Network (CFFN), to address the issue by integrating planar and stereo information of molecules in an interweaved fashion. The three-dimensional (3D, i.e., stereo) modality guarantees precision and completeness by providing unabridged information, while the two-dimensional (2D, i.e., planar) modality brings in chemical intuitions as prior knowledge for guidance. The zipper-like arrangement of 2D and 3D information processing promotes cooperativity between them, and their synergy is the key to our model's success. Experiments on various molecules or conformational datasets including a special newly created chiral molecule dataset comprised of various configurations and conformations demonstrate the superior performance of CFFN. The advantage of CFFN is even more significant in datasets made of small samples. Ablation experiments confirm that fusing 2D and 3D molecular graphs as unambiguous molecular descriptors can not only effectively distinguish molecules and their conformations, but also achieve more accurate and robust prediction of quantum chemical properties.

Identification of DNase I hypersensitive sites in the human genome by multiple sequence descriptors.

DNase I hypersensitive sites (DHSs) are chromatin regions highly sensitive to DNase I enzymes. Studying DHSs is crucial for understanding complex transcriptional regulation mechanisms and localizing cis-regulatory elements (CREs). Numerous studies have indicated that disease-related loci are often enriched in DHSs regions, underscoring the importance of identifying DHSs. Although wet experiments exist for DHSs identification, they are often labor-intensive. Therefore, there is a strong need to develop computational methods for this purpose. In this study, we used experimental data to construct a benchmark dataset. Seven feature extraction methods were employed to capture information about human DHSs. The F-score was applied to filter the features. By comparing the prediction performance of various classification algorithms through five-fold cross-validation, random forest was proposed to perform the final model construction. The model could produce an overall prediction accuracy of 0.859 with an AUC value of 0.837. We hope that this model can assist scholars conducting DNase research in identifying these sites.

A generic framework for hierarchical de novo protein design.

De novo protein design enables the exploration of novel sequences and structures absent from the natural protein universe. De novo design also stands as a stringent test for our understanding of the underlying physical principles of protein folding and may lead to the development of proteins with unmatched functional characteristics. The first fundamental challenge of de novo design is to devise "designable" structural templates leading to sequences that will adopt the predicted fold. Here, we built on the TopoBuilder (TB) de novo design method, to automatically assemble structural templates with native-like features starting from string descriptors that capture the overall topology of proteins. Our framework eliminates the dependency of hand-crafted and fold-specific rules through an iterative, data-driven approach that extracts geometrical parameters from structural tertiary motifs. We evaluated the TopoBuilder framework by designing sequences for a set of five protein folds and experimental characterization revealed that several sequences were folded and stable in solution. The TopoBuilder de novo design framework will be broadly useful to guide the generation of artificial proteins with customized geometries, enabling the exploration of the protein universe.

Evolving use of ancestry, ethnicity, and race in genetics research-A survey spanning seven decades.

To inform continuous and rigorous reflection about the description of human populations in genomics research, this study investigates the historical and contemporary use of the terms "ancestry," "ethnicity," "race," and other population labels in The American Journal of Human Genetics from 1949 to 2018. We characterize these terms' frequency of use and assess their odds of co-occurrence with a set of social and genetic topical terms. Throughout The Journal's 70-year history, "ancestry" and "ethnicity" have increased in use, appearing in 33% and 26% of articles in 2009-2018, while the use of "race" has decreased, occurring in 4% of articles in 2009-2018. Although its overall use has declined, the odds of "race" appearing in the presence of "ethnicity" has increased relative to the odds of occurring in its absence. Forms of population descriptors "Caucasian" and "Negro" have largely disappeared from The Journal (<1% of articles in 2009-2018). Conversely, the continental labels "African," "Asian," and "European" have increased in use and appear in 18%, 14%, and 42% of articles from 2009-2018, respectively. Decreasing uses of the terms "race," "Caucasian," and "Negro" are indicative of a transition away from the field's history of explicitly biological race science; at the same time, the increasing use of "ancestry," "ethnicity," and continental labels should serve to motivate ongoing reflection as the terminology used to describe genetic variation continues to evolve.

Characterization of indigenous populations of cannabis in Iran: a morphological and phenological study.

BACKGROUND: Cannabis is a historically, culturally, and economically significant crop in human societies, owing to its versatile applications in both industry and medicine. Over many years, native cannabis populations have acclimated to the various environments found throughout Iran, resulting in rich genetic and phenotypic diversity. Examining phenotypic diversity within and between indigenous populations is crucial for effective plant breeding programs. This study aimed to classify indigenous cannabis populations in Iran to meet the needs of breeders and breeding programs in developing new cultivars. RESULTS: Here, we assessed phenotypic diversity in 25 indigenous populations based on 12 phenological and 14 morphological traits in male and female plants. The extent of heritability for each parameter was estimated in both genders, and relationships between quantitative and time-based traits were explored. Principal component analysis (PCA) identified traits influencing population distinctions. Overall, populations were broadly classified into early, medium, and late flowering groups. The highest extent of heritability of phenological traits was found in Start Flower Formation Time in Individuals (SFFI) for females (0.91) Flowering Time 50% in Individuals (50% of bracts formed) (FT50I) for males (0.98). Populations IR7385 and IR2845 exhibited the highest commercial index (60%). Among male plants, the highest extent of Relative Growth Rate (RGR) was observed in the IR2845 population (0.122 g.g- 1.day- 1). Finally, populations were clustered into seven groups according to the morphological traits in female and male plants. CONCLUSIONS: Overall, significant phenotypic diversity was observed among indigenous populations, emphasizing the potential for various applications. Early-flowering populations, with their high RGR and Harvest Index (HI), were found as promising options for inclusion in breeding programs. The findings provide valuable insights into harnessing the genetic diversity of indigenous cannabis for diverse purposes.

Role of Hydrogen Bonding on the Design of New Hybrid Perovskites Unraveled by Machine Learning.

Selecting a set of reactants to accurately design a new low dimensional hybrid perovskite could greatly accelerate the discovery of materials with great potential in photovoltaics, or solid-state lighting. However, this design is challenging as most hybrid metal halides are not perovskites and no feature is clearly associated to the structural characteristics of the inorganic metal halide network. This work first demonstrates that the organic molecules are key parameters to determine the structure type of the inorganic network (i.e., perovskite versus non-perovskite). Then, machine learning (ML) algorithms are used to identify the key features of the organic cations leading to the perovskite structure type. Using a large dataset of hybrid metal halides, this work extracts the organic molecules of all hybrid lead halide compounds, calculates 2756 molecular descriptors and fingerprints for each of these molecules, and are able to predict through ML techniques if a specific organic amine will lead to the perovskite type with an accuracy up to 88.65%. Descriptors related to hydrogen bonding are identified as important features. Thus, a simple but reliable design principle could be demonstrated: the presence of primary ammonium cation is the primary condition to prepare hybrid lead halide perovskites regardless of their dimensionalities.

High-Throughput Calculations for Screening d- and p-Block Single-Atom Catalysts toward Li2 S/Na2 S Decomposition Guided by Facile Descriptor beyond Brønsted-Evans-Polanyi Relationship.

Single-atom catalysts (SACs) are promising cathode materials for addressing issues faced by lithium-sulfur batteries. Considering the ample chemical space of SACs, high-throughput calculations are efficient strategies for their rational design. However, the high throughput calculations are impeded by the time-consuming determination of the decomposition barrier (Eb ) of Li2 S. In this study, the effects of bond formation and breakage on the kinetics of SAC-catalyzed Li2 S decomposition with g-C3 N4 as the substrate are clarified. Furthermore, a new efficient and easily-obtained descriptor Li─S─Li angle (ALi─S─Li ) of adsorbed Li2 S, different from the widely accepted thermodynamic data for predicting Eb , which breaks the well-known Brønsted-Evans-Polanyi relationship, is identified. Under the guidance of ALi─S─Li , several superior SACs with d- and p-block metal centers supported by g-C3 N4 are screened to accelerate the sulfur redox reaction and fix the soluble lithium polysulfides. The newly identified descriptor of ALi─S─Li can be extended to rationally design SACs for Na─S batteries. This study opens a new pathway for tuning the performance of SACs to catalyze the decomposition of X2 S (X = Li, Na, and K) and thus accelerate the design of SACs for alkaline-chalcogenide batteries.

Chaperone Mimetic Strategy for Achieving Organic Room-Temperature Phosphorescence based on Confined Supramolecular Assembly.

The development of organic materials that deliver room-temperature phosphorescence (RTP) is highly interesting for potential applications such as anticounterfeiting, optoelectronic devices, and bioimaging. Herein, a molecular chaperone strategy for controlling isolated chromophores to achieve high-performance RTP is demonstrated. Systematic experiments coupled with theoretical evidence reveal that the host plays a similar role as a molecular chaperone that anchors the chromophores for limited nonradiative decay and directs the proper conformation of guests for enhanced intersystem crossing through noncovalent interactions. For deduction of structure-property relationships, various structure-related descriptors that correlate with the RTP performance are identified, thus offering the possibility to quantitatively design and predict the phosphorescent behaviors of these systems. Furthermore, application in thermal printing is well realized for these RTP materials. The present work discloses an effective strategy for efficient construction of organic RTP materials, delivering a modular model which is expected to help expand the diversity of desirable RTP systems.

Spectroscopic, computational, cytotoxicity, and docking studies of 6-bromobenzimidazole as anti-breast cancer agent.

6-Bromobenzimidazole (6BBZ) has been calculated in this study utilizing the 6-311++G(d,p) basis set and the Becke-3-Lee-Yang-Parr density functional approaches. The basic frequencies and geometric optimization are known. FTIR, FT-Raman, and UV-Vis spectra of the substance are compared between its computed and observed values. The energy gap between highest occupied molecular orbital-lowest unoccupied molecular orbital and molecule electrostatic potentials has been represented by charge density distributions that may be associated with the biological response. Time-dependent density functional theory calculations in the gas phase and dimethyl sulfoxide were carried out to ascertain the electronic properties and energy gap values using the same basis set. Molecular orbital contributions are investigated using the overlap population, partial, and total densities of states. Natural bond analysis was found to have strong electron delocalization by means of π(C4-C9) → π*(C5-C6), LP (N1) → π*(C7-C8), and LP(Br12) → π*(C5-C6) interactions. The Fukui function and Mulliken analysis have been explored on the atomic charges of the molecule. The nuclear magnetic resonance chemical shifts for 1 H and 13 C have been computed using the gauge-independent atomic orbital technique. With the highest binding affinity (-6.2 kcal mol-1 ) against estrogen sulfotransferase receptor (PDB ID: 1AQU) and low IC50 value of 17.23 µg/mL, 6BBZ demonstrated potent action against the MCF-7 breast cancer cell line. Studies on the antibacterial activity and ADMET prediction of the molecule have also been carried out.

MathFeature: feature extraction package for DNA, RNA and protein sequences based on mathematical descriptors.

One of the main challenges in applying machine learning algorithms to biological sequence data is how to numerically represent a sequence in a numeric input vector. Feature extraction techniques capable of extracting numerical information from biological sequences have been reported in the literature. However, many of these techniques are not available in existing packages, such as mathematical descriptors. This paper presents a new package, MathFeature, which implements mathematical descriptors able to extract relevant numerical information from biological sequences, i.e. DNA, RNA and proteins (prediction of structural features along the primary sequence of amino acids). MathFeature makes available 20 numerical feature extraction descriptors based on approaches found in the literature, e.g. multiple numeric mappings, genomic signal processing, chaos game theory, entropy and complex networks. MathFeature also allows the extraction of alternative features, complementing the existing packages. To ensure that our descriptors are robust and to assess their relevance, experimental results are presented in nine case studies. According to these results, the features extracted by MathFeature showed high performance (0.6350-0.9897, accuracy), both applying only mathematical descriptors, but also hybridization with well-known descriptors in the literature. Finally, through MathFeature, we overcame several studies in eight benchmark datasets, exemplifying the robustness and viability of the proposed package. MathFeature has advanced in the area by bringing descriptors not available in other packages, as well as allowing non-experts to use feature extraction techniques.

A merged molecular representation deep learning method for blood-brain barrier permeability prediction.

The ability of a compound to permeate across the blood-brain barrier (BBB) is a significant factor for central nervous system drug development. Thus, for speeding up the drug discovery process, it is crucial to perform high-throughput screenings to predict the BBB permeability of the candidate compounds. Although experimental methods are capable of determining BBB permeability, they are still cost-ineffective and time-consuming. To complement the shortcomings of existing methods, we present a deep learning-based multi-model framework model, called Deep-B3, to predict the BBB permeability of candidate compounds. In Deep-B3, the samples are encoded in three kinds of features, namely molecular descriptors and fingerprints, molecular graph and simplified molecular input line entry system (SMILES) text notation. The pre-trained models were built to extract latent features from the molecular graph and SMILES. These features depicted the compounds in terms of tabular data, image and text, respectively. The validation results yielded from the independent dataset demonstrated that the performance of Deep-B3 is superior to that of the state-of-the-art models. Hence, Deep-B3 holds the potential to become a useful tool for drug development. A freely available online web-server for Deep-B3 was established at http://cbcb.cdutcm.edu.cn/deepb3/, and the source code and dataset of Deep-B3 are available at https://github.com/GreatChenLab/Deep-B3.

Molecular properties of linear amino acids in water.

Four linear amino acids of increased separation of the carboxyl and amino groups, namely glycine (aminoacetic acid), ß-alanine (3-aminopropanoic acid), GABA (4-aminobutanoic acid) and DAVA (5-aminopentanoic acid), have been studied by quantum chemical ab initio and DFT methods including the solvent effect in order to get electronic structure and molecular descriptors, such as ionisation energy, electron affinity, molecular electronegativity, chemical hardness, electrophilicity index, dipole moment, quadrupole moment and dipole polarizability. Thermodynamic functions (zero-point energy, inner energy, enthalpy, entropy, and the Gibbs energy) were evaluated after the complete vibrational analysis at the true energy minimum provided by the full geometry optimization. Reaction Gibbs energy allows evaluating the absolute redox potentials on reduction and/or oxidation. The non-local non-additive molecular descriptors were compared along the series showing which of them behave as extensive, varying in match with the molar mass and/or separation of the carboxyl and amino groups. Amino acidic forms and zwitterionic forms of the substances were studied in parallel in order to compare their relative stability and redox properties. In total, 24 species were investigated by B3LYP/def2-TZVPD method (M1) including neutral molecules, molecular cations and molecular anions. For comparison, MP2/def2-TZVPD method (M2) with full geometry optimization and vibrational analysis in water has been applied for 12 species; analogously, for 24 substances, DLPNO-CCSD(T)/aug-cc-pVTZ method (M3) has been applied in the geometry obtained by MP2 and/or B3LYP. It was found that the absolute oxidation potential correlates with the adiabatic ionisation energy; the absolute reduction potential correlates with the adiabatic electron affinity and the electrophilicity index. In order to validate the used methodology with experimental vertical ionisation energies and vibrational spectrum obtained in gas phase, calculations were done also in vacuo.

Understanding the Influence of API Conformations on Amorphous Dispersion Formation Potential Predictions using the R3m Molecular Descriptor.

The R3m molecular descriptor (R-GETAWAY third-order autocorrelation index weighted by the atomic mass) has previously been shown to encode molecular attributes that appear to be physically and chemically relevant to grouping diverse active pharmaceutical ingredients (API) according to their potential to form persistent amorphous solid dispersions (ASDs) with polyvinylpyrrolidone-vinyl acetate copolymer (PVPVA). The initial R3m dispersibility model was built by using a single three-dimensional (3D) conformation for each drug molecule. Since molecules in the amorphous state will adopt a distribution of conformations, molecular dynamics simulations were performed to sample conformations that are probable in the amorphous form, which resulted in a distribution of R3m values for each API. Although different conformations displayed R3m values that differed by as much as 0.4, the median of each R3m distribution and the value predicted from the single 3D conformation were very similar for most structures studied. The variability in R3m resulting from the distribution of conformations was incorporated into a logistic regression model for the prediction of ASD formation in PVPVA, which resulted in a refinement of the classification boundary relative to the model that only incorporated a single conformation of each API.

From mundane to surprising nonadditivity: drivers and impact on ML models.

Nonadditivity (NA) in Structure-Activity and Structure-Property Relationship (SAR) data is a rare but very information rich phenomenon. It can indicate conformational flexibility, structural rearrangements, and errors in assay results and structural assignment. While purely ligand-based conformational causes of NA are rather well understood and mundane, other factors are less so and cause surprising NA that has a huge influence on SAR analysis and ML model performance. We here report a systematic analysis across a wide range of properties (20 on-target biological activities and 4 physicochemical ADME-related properties) to understand the frequency of various different phenomena that may lead to NA. A set of novel descriptors were developed to characterize double transformation cycles and identify trends in NA. Double transformation cycles were classified into "surprising" and "mundane" categories, with the majority being classed as mundane. We also examined commonalities among surprising cycles, finding LogP differences to have the most significant impact on NA. A distinct behavior of NA for on-target sets compared to ADME sets was observed. Finally, we show that machine learning models struggle with highly nonadditive data, indicating that a better understanding of NA is an important future research direction.

Machine-Learning-Assisted Descriptors Identification for Indoor Formaldehyde Oxidation Catalysts.

The development of highly efficient catalysts for formaldehyde (HCHO) oxidation is of significant interest for the improvement of indoor air quality. Up to 400 works relating to the catalytic oxidation of HCHO have been published to date; however, their analysis for collective inference through conventional literature search is still a challenging task. A machine learning (ML) framework was presented to predict catalyst performance from experimental descriptors based on an HCHO oxidation catalysts database. MnOx, CeO2, Co3O4, TiO2, FeOx, ZrO2, Al2O3, SiO2, and carbon-based catalysts with different promoters were compiled from the literature. Notably, 20 descriptors including reaction catalyst composition, reaction conditions, and catalyst physical properties were collected for data mining (2263 data points). Furthermore, the eXtreme Gradient Boosting algorithm was employed, which successfully predicted the conversion efficiency of HCHO with an R-square value of 0.81. Shapley additive analysis suggested Pt/MnO2 and Ag/Ce-Co3O4 exhibited excellent catalytic performance of HCHO oxidation based on the analysis of the entire database. Validated by experimental tests and theoretical simulations, the key descriptor identified by ML, i.e., the first promoter, was further described as metal-support interactions. This study highlights ML as a useful tool for database establishment and the catalyst rational design strategy based on the importance of analysis between experimental descriptors and the performance of complex catalytic systems.

Multi-Endpoint Acute Toxicity Assessment of Organic Compounds Using Large-Scale Machine Learning Modeling.

In recent years, alternative animal testing methods such as computational and machine learning approaches have become increasingly crucial for toxicity testing. However, the complexity and scarcity of available biomedical data challenge the development of predictive models. Combining nonlinear machine learning together with multicondition descriptors offers a solution for using data from various assays to create a robust model. This work applies multicondition descriptors (MCDs) to develop a QSTR (Quantitative Structure-Toxicity Relationship) model based on a large toxicity data set comprising more than 80,000 compounds and 59 different end points (122,572 data points). The prediction capabilities of developed single-task multi-end point machine learning models as well as a novel data analysis approach with the use of Convolutional Neural Networks (CNN) are discussed. The results show that using MCDs significantly improves the model and using them with CNN-1D yields the best result (R2train = 0.93, R2ext = 0.70). Several structural features showed a high level of contribution to the toxicity, including van der Waals surface area (VSA), number of nitrogen-containing fragments (nN+), presence of S-P fragments, ionization potential, and presence of C-N fragments. The developed models can be very useful tools to predict the toxicity of various compounds under different conditions, enabling quick toxicity assessment of new compounds.

Examining functional group-dependent effects on the ionization of lignin monomers using supercritical fluid chromatography/electrospray ionization mass spectrometry.

The chemical and biological conversion of biomass-derived lignin is a promising pathway for producing valuable low molecular weight aromatic chemicals, such as vanillin or guaiacol, known as lignin monomers (LMs). Various methods employing chromatography and electrospray ionization-mass spectrometry (ESI-MS) have been developed for LM analysis, but the impact of LM chemical properties on analytical performance remains unclear. This study systematically optimized ESI efficiency for 24 selected LMs, categorized by functionality. Fractional factorial designs were employed for each LM to assess ESI parameter effects on ionization efficiency using ultra-high-performance supercritical fluid chromatography/ESI-MS (UHPSFC/ESI-MS). Molecular descriptors were also investigated to explain variations in ESI parameter responses and chromatographic retention among the LMs. Structural differences among LMs led to complex optimal ESI settings. Notably, LMs with two methoxy groups benefited from higher gas and sheath gas temperatures, likely due to their lower log P and higher desolvation energy requirements. Similarly, vinyl acids and ketones showed advantages at elevated gas temperatures. The retention in UHPSFC using a diol stationary phase was correlated with the number of hydrogen bond donors. In summary, this study elucidates structural features influencing chromatographic retention and ESI efficiency in LMs. The findings can aid in developing analytical methods for specific technical lignins. However, the absence of an adequate number of LM standards limits the prediction of LM structures solely based on ESI performance data.

A systematic review of participant descriptors reported in studies of adhesive small bowel obstruction.

AIM: Reporting of participant descriptors in studies of adhesive small bowel obstruction (ASBO) can help identify characteristics associated with favourable outcomes and allow comparison with other studies and real-world clinical populations. The aim was to identify the pattern of participant descriptors reported in studies assessing interventions for ASBO. METHOD: This systematic review was registered with PROSPERO (CRD42021281031) and reported in line with the PRISMA checklist. Systematic searches of Ovid MEDLINE, Embase, and Cochrane Central Register of Controlled Trials (CENTRAL) were undertaken to identify studies assessing operative and non-operative interventions for adults with ASBO. Studies were dual screened for inclusion. Descriptors were categorised into conceptual domains by the research team. RESULTS: Searches identified 2648 studies, of which 73 were included. A total of 156 unique descriptors were identified. On average, studies reported 12 descriptors. The most frequently reported descriptors were sex, age, SBO aetiology, history of abdominal surgery, BMI and ASA classification. The highest number of descriptors in a single study was 34, compared to the lowest number of descriptors which was one. Pathway factors were the least frequently described domain. Overall, 37 descriptors were reported in just one study. CONCLUSION: There is a lack of consistency in participant descriptors reported in studies of SBO. Furthermore, a significant proportion of the descriptors were used infrequently. This makes it challenging to assess whether study participants are representative of the wider population. Further work is required to develop a Core Descriptor Set to standardise the reporting of patient characteristics and reduce heterogeneity between studies.

In silico prediction of ocular toxicity of compounds using explainable machine learning and deep learning approaches.

The accurate identification of chemicals with ocular toxicity is of paramount importance in health hazard assessment. In contemporary chemical toxicology, there is a growing emphasis on refining, reducing, and replacing animal testing in safety evaluations. Therefore, the development of robust computational tools is crucial for regulatory applications. The performance of predictive models is heavily reliant on the quality and quantity of data. In this investigation, we amalgamated the most extensive dataset (4901 compounds) sourced from governmental GHS-compliant databases and literature to develop binary classification models of chemical ocular toxicity. We employed 12 molecular representations in conjunction with six machine learning algorithms and two deep learning algorithms to create a series of binary classification models. The findings indicated that the deep learning method GCN outperformed the machine learning models in cross-validation, achieving an impressive AUC of 0.915. However, the top-performing machine learning model (RF-Descriptor) demonstrated excellent performance with an AUC of 0.869 on the test set and was therefore selected as the best model. To enhance model interpretability, we conducted the SHAP method and attention weights analysis. The two approaches offered visual depictions of the relevance of key descriptors and substructures in predicting ocular toxicity of chemicals. Thus, we successfully struck a delicate balance between data quality and model interpretability, rendering our model valuable for predicting and comprehending potential ocular-toxic compounds in the early stages of drug discovery.