Wisdom of Crowds for Supporting the Safety Evaluation of Nanomaterials - Data and results

<p>This repository contains the input data and the generated results utilized as part of the research article titled&nbsp;<em>"Wisdom of Crowds for Supporting the Safety Evaluation of Nanomaterials"</em>&nbsp; by Saarim&auml;ki &amp; Fratello et al.</p> <h1>Data</h1> <p>This dataset includes anonymized responses of a panel of experts to a questionnaire focused on nanomaterials safety.</p> <h2>Additional Data Sources</h2> <p>This entry further includes additional data from the following sources:</p> <ul> <li><strong>Saarim&auml;ki et al. (2021)</strong>: Preprocessed data and primary physicochemical characteristics are available on <a href="https://doi.org/10.5281/zenodo.6425445">Zenodo</a>.</li> <li><strong>Gallud et al. (2020)</strong>: Data is available from the NCBI Gene Expression Omnibus (GEO) under accession number <a href="https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE148705">GSE148705</a>.</li> <li><strong>del Giudice et al. (2023)</strong>: The advanced descriptors are accessible through the associated <a href="https://doi.org/10.5281/zenodo.7674574">Zenodo repository</a>.</li> <li><strong>Labouta et al. (2019)</strong>: The harmonized dataset of ENMs cell viability assays is available as <a href="https://pubs.acs.org/doi/suppl/10.1021/acsnano.8b07562/suppl_file/nn8b07562_si_001.xlsx">supporting information of the reasearch article</a>.</li> </ul> <h1>Outputs</h1> <ul> <li><strong>Statistical modeling:</strong> The inferred parameters of the statistical model developed to analyze the expert responses and identify the consensus among the experts.</li> <li><strong>Machine Learning classifiers:</strong> The performances of several machine learning classifiers trained on the consensus responses to learn models that can predict safety concerns of new nanomaterials based on transcriptomics and physicochemical descriptors data.</li> <li><strong>Feature Importance:</strong> The relevant features extracted from the models analyzed to understand which aspects of the molecular responses to ENMs and which physicochemical properties are most important in driving the predictions.</li> </ul> <h1>Contents</h1> <p><code>data/Combined_cleaned_responses_anon.xlsx</code>: Contains the anonymized responses from the expert panel.</p> <p><code>data/enms_grouping.txt</code>: Contains a categorization of the ENMs based on their core material.</p> <p><code>data/expert_bibliographies_anon.pickle</code>: Contains the anonymized bibliography of the experts to assess the multidisciplinarity of the panel assembled.</p> <p><code>data/gex.csc.gz</code>: Contains the gene expression after exposure to the ENMs.</p> <p><code>data/phenodata.txt</code>: Contains the meta-data of the experiments</p> <p><code>data/physicochemical_descriptors.txt</code>: Contains the physicochemical descriptors of the ENMs.</p> <p><code>data/external/nn8b07562_si_001.xlsx</code>: Contains a panel of harmonized ENMs cytotoxicity assays.</p> <p><code>outputs/concern_scores.csv</code>: Contains the consensus scores inferred by the statistical model.</p> <p><code>outputs/cross_validation/</code>: Contains the logs and performances of all the machine learning classification runs .</p> <p><code>outputs/important_{genes,physchem}_weighted.xlsx</code>: Feature relevance scores of both views.</p> <p><code>outputs/model_inference_anon.nc</code>: Inferred parameters of the statistical model.</p>