Find research datasets worth reusing

The code and data for the paper entitled 'Facilitating Efficient Discovery: A GUI-Oriented Approach for Exploring Functionality using Machine Learning Model'

<h2>Code</h2> dataCollection.py The collection of app data is primarily accomplished by processing and storing XML files and screenshots of the app. Firstly, the uiautomator is utilized to connect to the smart device and obtain screenshots and XML files. Subsequently, the XML files are analyzed to identify clickable functions within the GUI, and the textual information contained within the functions, along with their specific locations, is extracted. Finally, the Python file also includes handling of interface elements, such as determining if elements are obscured and verifying the legitimacy of the text. tagData.py The main implementation involves volunteers annotating app functionalities, including HTML generation, user data analysis, and retrieval. Flask framework is employed, presenting one GUI to the user each time while randomly prompting them to click on three functionalities. Ultimately, the time taken by users to locate these three functionalities is collected. userPersonalization.py Separating out the data annotated by each user facilitates personalized analysis. This process involves extraction, storage, and loading of individual user annotations. dataPreprocessing.py For a user-annotated functionality, completing the conversion from user time to either "hard-to-find" or "easy-to-find" involves several steps. First, the functionalities are vectorized, extracting relevant parameters from the XML files and computing their correlation with the time users spent searching for the functionalities. These parameters are then normalized to obtain feature vectors for the functionalities. Additionally, an initial determination is made regarding whether the annotated functionalities are "hard-to-find" or "easy-to-find" for each user. Subsequently, clustering is performed on all annotated data from users, and based on the clustering results, the outcomes are filtered and adjusted. difficultFindClassifier.py Train the classifier and use it to predict "hard-to-find" functionalities, then display the results. <h2>Data</h2> The data is located in the "static" folder: - The "persistentData" folder contains the trained classifier. - The "picture" folder contains screenshots of the app. - The "requestTime" folder stores data for when volunteers annotate only one function in a GUI. - The "threeResponseTime" folder saves data for when volunteers annotate three functions in a GUI. - The "userData" folder stores personalized user data. - The "xml_information" folder stores XML files of the app.

Tandem Mass Spectrometry Dataset for Machine Learning in Metabolomics

This dataset contains tandem mass spectrometry data cleaned and processed from the publicly available GNPS Spectral Library. We aim to continuously update this dataset with new data points as the spectral libraries expand.

openbsd-3-clauseMay 2024View details →

Stereoisomers are not Machine Learning's Best Friends: Experimental results of the prediction of the association constant between a cyclodextrin and a guest with Stereo2vec

This study addresses the challenge of accurately identifying stereoisomers in cheminformatics which originates from our objective to apply machine learning to predict association constant between a cyclodextrin and a guest. Identifying stereoisomers is indeed crucial for machine learning applications. Current tools offer various molecular descriptors, including their textual representation as Isomeric SMILES which can distinguish stereoisomers. But such representation is text-based and does not have a fixed size, so a conversion is needed to make it usable to machine learning approaches. Word embedding techniques can be used to solve this problem. Mol2vec, a word embedding approach for molecules, offers such a conversion. Unfortunately, it cannot distinguish between stereoisomers due to its inability to capture the spatial configuration of molecular structures. This study proposes several approaches that use word embedding techniques to handle molecular discrimination using stereochemical information of molecules or considering Isomeric SMILES notation as a text in Natural Language Processing. Our aim is to generate a distinct vector for each unique molecule, correctly identifying stereoisomer information in cheminformatics. The proposed approaches are then compared on our original machine learning task: predicting the association constant between a cyclodextrin and a guest molecule.

A Machine Learning Approach for Multiscale Modeling of Biological Tissues, Companion Dataset

This dataset includes the response of a Delaunay fiber network to various deformation gradients. The dataset is in adios2 format and includes stress, strain, strain energy, stiffness, and the applied deformation gradients. This also includes the pytorch implementation of our neural network.

Mental Health in Social Networks with Machine Learning Algorihtms

The DatasetMH.xlsx excel corresponds to a corpus of mental health in social networks labelled with polarity and stigma. In particular, the corpus consists of 2,287 comments labelled with polarity (positive, negative, neutral) and stigma from comments on Instagram posts about celebrity mental health disclosures: <ol> <li>Polarity: It consists of giving a positive, negative or neutral/undefined value to the comments in response to the disclosure or description of the symptomatology in the post. Positive polarity reflects understanding, encouragement or even admiration of the publication. E.g., “Cheer up, we love you".”. Negative polarity is assigned when the person expresses negative opinions, usually questioning the post with ironic, sarcastic or even mocking and disparaging comments. E.g., “how you show that you don't know what depression or anxiety is, shame on you!”. Neutral or undefined polarity is assigned in cases where no clear opinion is detected or can be interpreted in both directions. E.g., “take medication, it will help you" "and your partner?”</li> <li>Stigma: stigmatising responses to comments are behaviours in which negative beliefs and emotions towards MH problems are expressed. Stigma manifests in a variety of forms including rejection and anger against the person, which may extend to contempt or mockery, belittling their problem. E.g.,"What a desire to draw attention to yourself"; "what you have is a story"; "you're so inconsistent and seeking the limelight". Because socially we know that "stigma is wrong" many rejection comments are made in an ironic or sarcastic way. E.g., and how do you write on insta?"; "better information from someone who doesn't have a current account". Additionally, anger is shown by arguing that such posts "trivialise or commercialise" MH. E.g.,"don't come and tell me your false stories of overcoming, without even knowing what it is to work...". Other times the stigma manifests itself as pity or sorrow for the person. E.g.,“It breaks my heart”; “poor thing”.</li> </ol> The file DatasetMH_Emotions.xlsx corresponds to a corpus of mental health in social networks labelled with emotions. In particular, the corpus consists of 2,287 comments labelled with five emotions plus a neutral class from comments on Instagram posts about celebrity mental health disclosures. These emotions are: <ul> <li>Love/admiration: This emotion involves messages where admiration, approval and love are closely related.</li> <li>Gratitude: the messages imply a sincere appreciation for sharing mental health-related content on social networks.</li> <li>Comprehension/empathy/identification: The messages involve interest in and understanding of the message, including self-identification with the situation or context.</li> <li>Sadness: This primary emotion is produced by events that are not pleasant and that denote heaviness. It includes many manifestations of pity for the person.</li> <li>Anger/contempt/mockery: This emotion involves responses of irritation and attacks on the person as ridiculous and superficial.</li> <li>Neutral: This category corresponds to messages without emotions.</li> </ul> The labelling process of both datasets was divided into two phases: an initial phase with a pilot corpus (N = 787 comments) and a second phase focused on the development of the corpus with all the comments of the selected posts (N = 21151). The same methodology was followed in both phases: once the comments were collected, the corpus was cleaned, and then two independent experts were responsible for labelling each category. A third expert then reviewed the comments to resolve discrepancies. In the third and final phase, a final corpus for application to the machine learning algorithms is built from the large corpus (N = 2287). Classification models are a set of machine learning algorithms developed to assess emotional response, i.e. polarity, stigma and emotions in social networks, based on previously developed datasets (DatasetMH_Emotions.xlsx, DatasetMH.xlsx).

opencc-by-4.0Dec 2023View details →

aurora: A Machine Learning GWAS Tool For Analyzing Microbial Habitat Adaptation

This upload contains compressed raw data to rerun examples in the main text of aurora and the first version of aurora used in the paper.

Data for the paper "PETScML: Second-Order Solvers for Training Regression Problems in Scientific Machine Learning"

Open the record for dataset details and reuse information.

opencc-zeroMay 2024View details →

DFT torsiondrive data for: MACE-OFF23: Transferable Machine Learning Force Fields for Organic Molecules

MACE-OFF23: Transferable Machine Learning Force Fields for Organic Molecules <div><a href="https://arxiv.org/search/physics?searchtype=author&query=Kov%C3%A1cs,+D+P">Dávid Péter Kovács</a>, <a href="https://arxiv.org/search/physics?searchtype=author&query=Moore,+J+H">J. Harry Moore</a>, <a href="https://arxiv.org/search/physics?searchtype=author&query=Browning,+N+J">Nicholas J. Browning</a>, <a href="https://arxiv.org/search/physics?searchtype=author&query=Batatia,+I">Ilyes Batatia</a>, <a href="https://arxiv.org/search/physics?searchtype=author&query=Horton,+J+T">Joshua T. Horton</a>, <a href="https://arxiv.org/search/physics?searchtype=author&query=Kapil,+V">Venkat Kapil</a>, <a href="https://arxiv.org/search/physics?searchtype=author&query=Witt,+W+C">William C. Witt</a>, <a href="https://arxiv.org/search/physics?searchtype=author&query=Magd%C4%83u,+I">Ioan-Bogdan Magdău</a>, <a href="https://arxiv.org/search/physics?searchtype=author&query=Cole,+D+J">Daniel J. Cole</a>, <a href="https://arxiv.org/search/physics?searchtype=author&query=Cs%C3%A1nyi,+G">Gábor Csányi </a><a href="https://doi.org/10.48550/arXiv.2312.15211">https://doi.org/10.48550/arXiv.2312.15211</a></div>   Supporting data including raw outputs from SPICE consistent torsion drives on the TorsionNet500 and OpenFF Biaryl datasets and HDF5 versions formated to be consistent with the rest of the SPICE dataset. See the <a href="../records/10975225">SPICE release</a> for more details.

opencc-by-4.0Dec 2023View details →

SonicParanoid2: fast, accurate, and comprehensive orthology inference with machine learning and language models

This repository contains the documentation, test datasets and scripts used in the following study: "SonicParanoid2: fast, accurate, and comprehensive orthology inference with machine learning and language models" - `sonic-manuscript-master.zip` contains all the scripts to reproduce the study, including those for generating the figures and tables included in the manuscript. - <a href="../api/records/11361985/draft/files/sonicparanoid2.wiki.tar.xz/content" target="_blank" rel="noopener noreferrer">sonicparanoid2.wiki.tar.xz</a> contains a snapshot fo the wiki for SonicParanoid2 as of May 30, 2024

openmit-licenseMay 2024View details →

Machine learning based on whole-rock geochemical data: An indication for the porphyry Cu-Au deposits in Central Asian Orogenic Belt

<div> <h2>Introduction</h2> </div> Welcome to the PorphyryAuML project! This repository is dedicated to exploring the mechanism of gold (Au) enrichment in porphyry systems within the Central Asian Orogenic Belt using machine learning. We apply models like XGBoost and Random Forest to analyze whole-rock geochemical data, aiming to classify porphyry deposit types and highlight key geochemical indicators. <div> <h2>Key Features 🌟</h2> </div> <ul> <li>Principal Component Analysis (PCA): Reduce dimensionality to discover the most significant variables.</li> <li>Machine Learning Models: Utilize XGBoost and Random Forest for robust classification.</li> <li>Feature Importance Analysis: Identify crucial geochemical markers for Au presence and quantity.</li> <li>Visualization: Detailed plots to illustrate model outcomes and geochemical patterns.</li> </ul> <div> <h2>Data 📊</h2> </div> <div> <h3>DATA.xlsx</h3> </div> The <code>DATA.xlsx</code> file contains crucial geochemical data used in our analysis, organized across three sheets: <ul> <li>group1: Represents the Cu-Au (Copper-Gold) porphyry deposits. This sheet contains all relevant geochemical markers and measurements specific to this group.</li> <li>group2: Corresponds to Cu(-Au±Mo) (Copper with minor Gold and possibly Molybdenum) porphyry deposits. It includes a detailed set of data focusing on the variations and characteristics of these mixed element deposits.</li> <li>group3: Contains data related to Cu-Mo (Copper-Molybdenum) porphyry deposits, focusing on the distinct geochemical profiles that typify these deposits.</li> </ul> Each sheet is named to reflect the group it represents and is vital for our machine learning analysis to classify and predict porphyry deposit types based on their geochemical properties.

RepliChrom: Interpretable machine learning predicts cancer-associated enhancer-promoter interactions using DNA replication timing

This dataset accompanies the study "RepliChrom: Interpretable machine learning predicts cancer-associated enhancer-promoter interactions using DNA replication timing". The study introduces RepliChrom, a computational framework designed to predict enhancer–promoter interactions (EPIs) by leveraging multi-scale replication timing (RT) signals. This approach addresses the fundamental challenge of distinguishing gene targets regulated by distal enhancers from those activated by proximal transcriptional activity-a key problem in understanding the causal basis of complex diseases. Despite recent advances in high-throughput technologies such as Hi-C, ChIA-PET, and Hi-TrAC that allow genome-wide reconstruction of 3D chromatin architecture, the role of DNA replication timing in mediating these spatial interactions remains underexplored. RepliChrom fills this gap by using cell-type-specific RT profiles as predictive features for chromatin interaction inference. To support model development, training, and evaluation, we provide a comprehensive multi-omics dataset covering six human cell lines (K562, GM12878, HeLaS3, IMR90, NHEK, and HUVEC ), encompassing: Hi-C datasets: Processed chromatin interaction loops used to define positive and negative enhancer–promoter interaction pairs. ChIA-PET datasets: Interaction data anchored around transcription factor binding, including POLR2A and CTCF, used for model validation across different interaction types. Hi-TrAC datasets: Targeted chromatin accessibility-derived interaction data, offering complementary validation of the model on alternate platforms. Replication Timing (RT) data: Processed RT signal profiles for each cell line, used to extract multi-scale temporal features as inputs for RepliChrom. These datasets enable reproducibility of the model training process and serve as benchmark resources for future research into DNA replication–mediated regulation of 3D genome architecture. Included Files Hi-C_datasets.zip (2.23 MB): Processed Hi-C interaction pairs for six cell types. ChIA-PET_datasets.zip (818.18 KB): CTCF and POLR2A ChIA-PET interactions across multiple lines. Hi-TrAC_datasets.zip (196 bytes): Hi-TrAC-based chromatin interaction training datasets across multiple lines. Cellline_RT_data.zip (86.17 MB): Replication timing signal data across multiple human cell types for multi-scale replication timing feature extraction. Usage All datasets are intended for academic, non-commercial use. The provided files can be directly used to reproduce the training and evaluation of RepliChrom, and may also support broader applications in enhancer–promoter modeling, replication-timing analysis, and 3D genomics studies. For detailed usage instructions and code implementation, please refer to the GitHub repository: https://github.com/DaoFuying/RepliChrom

Machine-Learning Based Location of the 2021 MW 7.4 Maduo Earthquake Sequence: Insight into Intraplate Seismogenesis

This file is the Machine-Learning Based earthquake catalog of the 2021 MW 7.4 Maduo Earthquake. It is only used for scientific research. ATTENTION!!! The data for the paper "Relocation of the 2024 MS 7.1 Wushi, Xinjiang earthquake sequence and implications for seismogenic structure" is accessible at the website "https://zenodo.org/records/12790377". ATTENTION!!!

Machine learning-based pulse wave analysis for classification of circle of Willis topology: an in silico study with 30,618 virtual subjects (database: Complete CoW)

This repository contains the dataset for the complete CoW described in the article with the same name. MATLAB and Python codes for post-processing the dataset and the code for training and testing all machine learning models using the open-source library TensorFlow 2.12, the Keras application programming interface, and the Scikit-learn Python package can be found in here (<a href="https://zenodo.org/records/12519322" target="_blank" rel="noopener">https://zenodo.org/records/12519322</a>).

Governance Opportunities in the Machine Learning Pipeline

This framework maps the technical ML pipeline alongside key governance questions and decisions that impact and can benefit from the input from wider communities.

openapache2.0Jun 2024View details →

Automated Trustworthiness Testing for Machine Learning Classifiers

This repository includes data for the paper Automated Trustworthiness Testing for Machine Learning Classifiers.

dryad36/100

Data from: OoCount: A machine-learning based approach to mouse ovarian follicle counting and classification

The number and distribution of ovarian follicles in each growth stage provides a reliable readout of ovarian health and function. Leveraging techniques for three-dimensional (3D) imaging of ovaries in toto has the potential to uncover total, accurate ovarian follicle counts. However, because of the size and holistic nature of these images, counting oocytes is time consuming and difficult. The advent of deep-learning algorithms has allowed for the rapid development of ultra-fast, automated methods to analyze microscopy images. In recent years, these pipelines have become more user-friendly and accessible to non-specialists. We used these tools to create OoCount, a high-throughput, open-source method for automatic oocyte segmentation and classification from fluorescent 3D microscopy images of whole mouse ovaries using a deep-learning convolutional neural network (CNN) based approach. We developed a fast clearing and spinning disk confocal-based imaging protocol to obtain 3D images of whole mount perinatal and adult mouse ovaries. Then, fluorescently labeled oocytes from 3D images of ovaries were manually annotated to develop a machine learning training dataset. This dataset was used to train a CNN to automatically label all oocytes in the ovary. In a second phase, we trained another CNN to classify labeled oocytes and sort them into growth stages. Using OoCount, we can obtain accurate counts of oocytes in each growth stage in the perinatal and adult ovary, improving our ability to study ovarian function and fertility. Here, we provide an end-to-end protocol for developing high quality 3D images of the perinatal and adult mouse ovary, obtaining follicle counts and stages, and how to customize OoCount to fit images produced in any lab.

opencc-zeroJun 2024View details →

Weather data (forecast and observation) at 48 locations in France for beginning of 2024 for Machine Learning Training

The data provided data are historical weather measurement and forecast at 48 locations in France and its boundary. Measurements are inside files named MES_YYYY.csv with YYYY is the id code of the station. The file "Station_list.csv" contains the list of the 45 locations with the id code, the name and then the latitude and longitude. Forecasts are inside files named XXX_YYYY.csv with YYYY the id code corresponding of the location of the grid ouput close to the associated observation location. XXX is the id of the numerical forecast:     "GFS0.25-Complet" for GFS file at 0.25° resolution     "LEXIS" for WRF produced by EVEREST project using the LEXIS chain     "WRF3KM-Complet" for WRF at 3km resolution produced by NUMTECH     "WRF12KM-Complet" for WRF at 12km resolution produced by NUMTECH Description of MES-YYYY files: - One line per measurement with hourly resolution - columns are: Date(TU),Temperature2m_degC,WindSpeed10m_m/s,WindDirection10m_m/s     Date = date of measurement in TU and format DD/MM/YYYY HH:MM     Temperature2m_degC = air temperature at 2m height in °Celsius     WindSpeed10m_m/s = wind speed at 10m height in m/s     WindDirection10m_deg = wind direction at 10m height in deg. (0 or 360 = wind from north to south, 45°=wind from east to east, ....) If measurement is not available for a specific hour for one parameter, the value "-999" is used. The observation data gocfrom 28/01/2024 00HTU to 17/03/2024 23HTU Description of XXX_YYYY forecast files: - One line per forecast with hourly resolution - columns are: First date run (TU),Forecast date,Temperature2m_degC,WindSpeed10m_m/s,WindDirection10m_m/s     First date run (TU) = date of start of the forecast in TU and format DD/MM/YYYY HH:MM. HH could be 00 and 12 according to the cycle of forecast start.     Forecast date = date of the forecast in TU and format DD/MM/YYYY HH:MM. HH go from 00 to 23.      Temperature2m_degC = air temperature at 2m height in °Celsius     WindSpeed10m_m/s = wind speed at 10m height in m/s     WindDirection10m_deg = wind direction at 10m height in deg. (0 or 360 = wind from north to south, 45°=wind from east to east, ....) If forecast is not available for a specific hour for one parameter, the value "-999" is used. The forecast data go from 28/01/2024 00HTU to 17/03/2024 23HTU

opencc-by-4.0Dec 2023View details →

Inferring Surface NO2 over Western Europe: A Machine Learning Approach with Uncertainty Quantification

The data that serves to substantiate the analysis presented in the article.

Cloud-resolving model for machine learning buoyant cloudy updraught

Atmospheric model data from simulations carried out using the Unified Model nesting suite u-cj161. This repository includes scripts to perform all the required steps. Retrieval of the model output files from the archive: ml_cape_pdf_1_retrieve_lam_files.py  Coarse-graining of the 1.5 km atmospheric model data to a series of different "global model" resolutions by calculating mean profiles of key thermodynamic variables and the calculation of the fraction of that coarse-volume in which the 1.5 km pixels meet the criterion of being bouyant cloudy updraughts: ml_cape_pdf_2d_coarse_grain_bcu.py  Subsampling to rebalance the data set since so much of the raw data consists of examples of no convectie activity: ml_cape_pdf_3e_sub_sample_bcu.py  There was too much data to process all the raw data in one go, so it was done is several steps. There is then a need to join multiple files together: ml_cape_pdf_3g_stitch_together_bcu_files.py  The data is on the Unified Model L70 grid (counting from the bottom up), but the goal is to deploy the machine-learnt algorithm on the E3SM grid (72 levels counting from the top down), so some regirddign is required: ml_cape_pdf_3h_regrid_um_to_e3sm.py  Training of 1 two-headed 1d CNN is done using: ml_cape_pdf_4c_bcu_cnn.py  Evaluation against witheld data is done using: ml_cape_pdf_5b_validate_BCU.py

Addressing Observational Gaps in Aerosol Parameters using Machine Learning: Implications to Aerosol Radiative Forcing

This dataset represents Aerosol Optical Depth (AOD), Single Scattering Albedo (SSA), and Absorption Parameter (AP) data over Kanpur, India, sourced from AERONET with initial data gaps of approximately 37%, 62%, and 58% respectively. To reduce these gaps, XGBoost, a machine learning model trained with reanalysis and satellite datasets, was employed with optimized hyperparameter tuning. Using AERONET data for training, XGBoost effectively addressed gaps, improving AOD by 10%, SSA by 23%, and AP by 21%.