Find research datasets worth reusing
Search datasets from major research repositories and use ShareScore to quickly assess how well each record supports discovery, access, and reuse.
125
datasets available to search
ShareScore release 0.9.0
Dataset results
125 results for “Structure learning”
Bioactivity deep learning for structure-free compound-protein interaction
<p>CPI2M data for "<strong>Bioactivity deep learning for structure-free compound-protein interaction</strong>".</p> <p>CPI2M_main_Ki.csv: Bioactivity data with <strong>pKi </strong>activity type. Used for model training and internal validation.</p> <p>CPI2M_main_Kd.csv: Bioactivity data with <strong>pKd</strong> activity type. Used for model training and internal validation.</p> <p>CPI2M_main_EC50.csv: Bioactivity data with <strong>pEC50 </strong>activity type. Used for model training and internal validation.</p> <p>CPI2M_main_IC50.csv: Bioactivity data with <strong>pIC50 </strong>activity type. Used for model training and internal validation.</p> <p>CPI2M_few_Ki.csv: Bioactivity data with <strong>pKi </strong>activity type. Used for external validation.</p> <p>CPI2M_few_Kd.csv: Bioactivity data with <strong>pKd </strong>activity type. Used for external validation.</p> <p>CPI2M_few_EC50.csv: Bioactivity data with <strong>pEC50 </strong>activity type. Used for external validation.</p> <p>CPI2M_few_IC50.csv: Bioactivity data with <strong>pIC50 </strong>activity type. Used for external validation.</p> <p>potency.csv: BIoactivity data with <strong>pPotency </strong>activity type. Not used currently but can be potentially adopted as classification data for customized use.</p> <p>percentage.csv: BIoactivity data with <strong>Percentage Inhibition </strong>activity type. Not used currently but can be potentially adopted as classification data for customized use.</p> <p>Protein_pretrained_feat.zip: pre-calculated protein feature files with UniProt ID naming. <strong>Should be unzipped</strong> before start model training with CPI2M data.</p> <p> </p> <p>For each .csv data, columns include "<strong>smiles</strong>" (ligand SMILES), "<strong>exp_mean</strong>" (nM bioactivity), "<strong>y</strong>" (neg.log nM, final label), "<strong>cliff_mol</strong>" (whether activity cliff or not), "<strong>split</strong>" (splitting label by activity cliff), "<strong>Uniprot_id</strong>" (UniProt ID for protein), "<strong>Sequence</strong>" (wildtype sequence for protein), and "type_id" (bioactivity type token, pKi =0, pKd=1, pEC50=2, pIC50=3).</p> <p> </p> <p>Please find the project code at https://github.com/gu-yaowen/GGAP-CPI</p> <p> </p>
Docked structures from "Optimizing active learning for free energy calculations"
<p>This archive contains the docked TYK2 structures used in the paper "Optimizing active learning for free energy calculations" (<a title="Persistent link using digital object identifier" href="https://doi.org/10.1016/j.ailsci.2022.100050" target="_blank" rel="noreferrer noopener"><span><span>https://doi.org/10.1016/j.ailsci.2022.100050</span></span></a>). AM1-BCC charges are stored in the field "AM1Cache" in the SD file. The charges can be extracted using the code sample below. </p> <p> </p> <pre><code>from rdkit import Chem import base64 import pickle suppl = Chem.SDMolSupplier("10k_most_similar_tyk2_charged.sdf", removeHs=False) for mol in suppl: am1 = mol.GetProp("AM1Cache") am1_charges = pickle.loads(base64.b64decode(mol.GetProp("AM1Cache"))) assert len(am1_charges) == mol.GetNumAtoms(), "Charge cache has different number of charges than mol atoms"</code></pre>
Unstable Crystallographic & Molecular Structures for Machine Learning of System Energies
<div> <div> <div> <p>Extended QM9 (E-QM9) includes diverse sizes (i.e. number of atoms) and compositions of OoE molecules, through extending a subset of QM9 with OoE versions of 10k of its molecules.</p> <p>Periodic crystals (PC) allows learning regular bonding patterns that arise in periodic structures by repeating the base crystal lattice. We use the Face-Centred Cubic (fcc) Bravais lattice for aluminium (Al) and copper (Cu) crystals.</p> <p>Crystal Growth (CG) contains growing crystals of increasing size and complexity. Starting from a basic fcc crystal seed of 14 atoms, new systems are generated by iteratively placing atoms at a random location on the surface of the growing crystal following its lattice pattern, with sizes ranging from 15 to 114 atoms. We use 20 random seeds for each atom type, thus creating 40 varied Al and Cu crystal growths and 4,000 stable systems. As a result, for a given crystal size and composition (atom type), there are 20 samples with differently located atoms. CG enables experi- menting with large scale atomic interactions in non-regular sys- tems, and enables evaluation of an ML method’s ability to learn how each atom contributes to the final potential energy.</p> <p>In all datasets, OoE systems are obtained by compressing/dilating all interatomic distances (i.e. isometrically) at regular intervals within 90-150% of stable geometry, which we refer to as ‘scaling’. In other words, scaling is applied to the coordinates of all atoms within the system. At each geometry, the ground-truth potential energy is calculated using CP2K7’s DFT.</p> </div> </div> </div>
Project files provided as supporting information to the manuscript "A deep learning approach to the structural analysis of proteins"
<p><strong>README file to the project files provided as supporting information to the manuscript “A deep learning approach to the structural analysis of proteins”</strong></p> <p>Dec. 30, 2018</p> <p>Authors: Marco Giulini and Raffaello Potestio</p> <p>==================================</p> <p>The dataset contains the following files:</p> <p> </p> <p>- datasets.zip: archive containing five .csv files, namely:</p> <p> - decoys_cm.csv : all the data for 10728 protein decoys, training set</p> <p> - evaluation_cm.csv : all data for 146 proteins in the evaluation set</p> <p> - random_CG.csv : 1200 Coulomb matrices. 100 CG models for each protein with 120 amino acids</p> <p> - 1e5g_centered_sphere.csv : 100 CG models in which the central atoms in 1e5g are not removed</p> <p> - 1e5g_random_sphere.csv : 10 CG models for 10 different (random) locations for the sphere that includes atoms that have to be retained. 100 CG models in total</p> <p> </p> <p>- decoys_labels.lab containing the labels associated to the 10728 decoys present in the training set</p> <p>- evaluation_labels.lab containing the labels associated to the 146 pdb files in the evaluation set</p> <p>- random_CG_labels.lab containing the labels associated to the 6 proteins with 120 amino acids</p> <p>- network_development_training: a python script that performs cross validation and full training of the model</p> <p>- saved_networks.zip FOLDER containing 10 networks: the architecture is included in .json files while weight parameters are inside .hs files</p> <p> </p> <p>- pdb_files.zip FOLDER containing the PDB files that have been employed in the project, namely:</p> <p> - pdb_files_len100 : pdb files with 100 amino acids</p> <p> - pdb_files_len101-110 : pdb files with a number of amino acids between 101 and 110</p> <p> - decoys : decoys of length 100 extracted from the above folder: name syntax == PDBNAME_decoy_STARTRES_ENDRES.pdb</p> <p> EXAMPLE 6gsp.pdb will give rise to 6gsp_decoy_0_100.pdb , 6gsp_decoy_1_101.pdb , 6gsp_decoy_2_102.pdb , 6gsp_decoy_3_103.pdb , 6gsp_decoy_4_104.pdb</p> <p> - pdb_files_len100 : 6 pdb files with 120 amino acids</p> <p> </p>
Structured Power Grid Simulation Dataset for Machine Learning: Failure and Survival Events in Grid2Op's L2RPN WCCI 2022 Environment
<p>This dataset was developed for and used in the paper titled <em>"Fault Detection for Agents in Power Grid Topology Optimization: A Comprehensive Analysis"</em> by Malte Lehna, Mohamed Hassouna, Dmitry Degtyar, Sven Tomforde, and Christoph Scholz, presented at the <em>Workshop on Machine Learning for Sustainable Power Systems (ML4SPS)</em>, part of <em>ECML PKDD 2024</em>. While the paper is pending formal publication, a preprint version is available on arXiv.</p> <p>The dataset contains structured training, validation, and test data comprising failure and survival events observed in transmission power grid simulations. These were generated using Grid2Op with the WCCI 2022 L2RPN environment. Each data instance is labeled with one of four classes, representing survival or impending failure in 1, 3, and 5 timesteps. This dataset was used to train, validate and test machine learning models that predict grid agent failures in topology optimization tasks. </p>
Structural Interaction Fingerprints and Machine Learning for predicting and explaining binding of small molecule ligands to RNA: a benchmark dataset
<p><b>Structural Interaction Fingerprints and Machine Learning for predicting and explaining binding of small molecule ligands to RNA: a benchmark dataset.</b></p><p>Ribonucleic acids (RNA) play crucial roles in living organisms as they are involved in key processes necessary for proper cell functioning. Some RNA molecules, such as bacterial ribosomes and precursor messenger RNA, are targets of small molecule drugs, while others, e.g., bacterial riboswitches or viral RNA motifs are considered as potential therapeutic targets. Thus, the continuous discovery of new functional RNA increases the demand for developing compounds targeting them and for methods for analyzing RNA—small molecule interactions. We recently developed fingeRNAt - a software for detecting non-covalent bonds formed within complexes of nucleic acids with different types of ligands. The program detects several non-covalent interactions, such as hydrogen and halogen bonds, ionic, Pi, inorganic ion- and water-mediated, lipophilic interactions, and encodes them as computational-friendly Structural Interaction Fingerprint (SIFt). Here we present the application of SIFts accompanied by machine learning methods for binding prediction of small molecules to RNA targets. We show that SIFt-based models outperform the classic, general-purpose scoring functions in virtual screening. We discuss the aid offered by Explainable Artificial Intelligence in the analysis of the binding prediction models, elucidating the decision-making process, and deciphering molecular recognition processes.</p>
MISATO - Machine learning dataset for structure-based drug discovery
<p>Developments in Artificial Intelligence (AI) have had an enormous impact on scientific research in recent years. Yet, relatively few robust methods have been reported in the field of structure-based drug discovery. To train AI models to abstract from structural data, highly curated and precise biomolecule-ligand interaction datasets are urgently needed. We present MISATO, a curated dataset of almost 20000 experimental structures of protein-ligand complexes, associated molecular dynamics traces, and electronic properties. Semi-empirical quantum mechanics was used to systematically refine protonation states of proteins and small molecule ligands. Molecular dynamics traces for protein-ligand complexes were obtained in explicit water. The dataset is made readily available to the scientific community via simple python data-loaders. AI baseline models are provided for dynamical and electronic properties. This highly curated dataset is expected to enable the next-generation of AI models for structure-based drug discovery. Our vision is to make MISATO the first step of a vibrant community project for the development of powerful AI-based drug discovery tools.</p>
Learning Naturalistic Temporal Structure in the Posterior Medial Network
Open the record for dataset details and reuse information.
Research data supporting: "Machine learning of microscopic structure-dynamics relationships in complex molecular systems"
<p>This repository contains the set of data and the code to reproduce the results shown in "Machine learning of microscopic structure-dynamics relationships in complex molecular systems" published on Machine Learning: Science and Technology (DOI: 10.1088/2632-2153/ad0fa5).</p>
Small molecules targeting the structural dynamics of AR-V7 partially disordered protein using deep learning and physics based models.
<p>Partially disordered proteins can contain both stable and unstable secondary structure segments and are involved in various (mis)functions in the cell. The extensive conformational dynamics of partially disordered proteins scaling with extent of disorder and length of the protein hampers the efficiency of traditional experimental and in-silico structure-based drug discovery approaches. Therefore new efficient paradigms in drug discovery taking into account conformational ensembles of proteins need to emerge. In this study, using as a test case the AR-V7 transcription factor splicing variant related to prostate cancer, we present an automated methodology that can accelerate the screening of small molecule binders targeting partially disordered proteins. By swiftly identifying the conformational ensemble of AR-V7, and reducing the dimension of binding-sites by a factor of 90 by applying appropriate physicochemical filters, we combine physics based molecular docking and multi-objective classification machine learning models that speed up the screening of thousands of compounds targeting AR-V7 multiple binding sites. Our method not only identifies previously known binding sites of AR-V7, but also discovers new ones, as well as increases the multi-binding site hit-rate of small molecules by a factor of 17 compared to naive physics-based molecular docking. </p>
Predicting glycan structure from tandem mass spectrometry via deep learning
<p>Curated set of LC-MS/MS data from glycomics studies. Used for training and applying CandyCrunch, a deep learning model to predict glycan structure from LC-MS/MS data, described in Urban et al., Nat Methods, 2024 and https://github.com/BojarLab/CandyCrunch.</p> <p>Files:</p> <p>full_dataset.xlsx: Full dataset with all annotated LC-MS/MS glycan spectra</p> <p>X_train.pkl: spectra and metadata from our training set</p> <p>y_train.pkl: labels from our training set</p> <p>X_test.pkl: spectra and metadata from our independent test set</p> <p>y_test.pkl: labels from our independent test set</p> <p>glycans.pkl: glycans in IUPAC-condensed nomenclature in the same order as the label-encoding</p>
Capturing functional relations in fluid-structure interaction via machine learning
<p>While fluid-structure interaction (FSI) problems are ubiquitous in various applications from cell-biology to aerodynamics, they involve huge computational overhead. In this paper, we adopt a machine learning (ML)-based strategy to bypass the detailed FSI analysis that requires cumbersome simulations in solving the Navier-Stokes (N-S) equations. To mimic the effect of fluid on an immersed beam, we have introduced dissipation into the beam model with time-varying forces acting on it. The forces in a discretized setup have been decoupled via an appropriate linear algebraic operation, which generates the ground truth force/moment data for the ML analysis. The adopted ML technique, symbolic regression, generates computationally tractable functional forms to represent the force/moment with respect to space and time. These estimates are fed into the dissipative beam model to generate the immersed beam's deflections over time, which are in conformity with the detailed FSI solutions. Numerical results demonstrate that the ML-estimated continuous force and moment functions are able to accurately predict the beam deflections under different discretizations.</p>
Source codes and datasets for the paper "DRLComplex: Reconstruction of protein quaternary structures using deep reinforcement learning"
<p>This contains the<strong> reproducible source code and dataset </strong>for the paper "DRLComplex : Reconstruction of protein quaternary structures using deep reinforcement learning paper"</p>
Pose Selector Workflow - Structure Input Files for Machine Learning (Set 2)
<p>Second set of structure input files for the docking poses of the remaining 2022 protein-ligand complexes. Together with the structure files and absolute binding free energy (ABFE) estimates shared in 10.5281/zenodo.11397017, this data can be used to train a machine-learning (ML) model predicting the ABFE of binding poses of protein-ligand complexes.</p> <p>More background on the workflow generating the structure files and ABFE estimates is provided in 10.5281/zenodo.11397017.</p>
Pose Selector Workflow - Docking Poses, Absolute Binding Free Energy Estimates and Structure Input Files for Machine Learning
<p>The Pose Selector (PS) workflow calculates absolute binding free energies (ABFEs) for binding poses of protein-ligand complexes. First, it converts the binding poses (both docking poses as well as experimentally observed ligand binding poses), which are provided as a combination of protein PDB file and ligand MOL2 file, into input files for molecular dynamics (MD) simulations with GROMACS after they have passed extensive quality checks and repair steps. Next, the PS workflow post-processes and analyses the last frame of the resulting eight 100 ps trajectories per binding pose with the Generalised Born model of implicit solvation as implemented in gmx_MMPBSA to obtain the ABFE estimates. The workflow was designed for soluble proteins without post-translational modifications, co-factors and non-standard amino acids, and it has limited support for coordinated ions.</p> <p>For the dataset published here, the PS workflow was run on docking poses generated for the PDBbind 2020 dataset (http://www.pdbbind.org.cn/index.php), shared in dockingPosesPDBBind2020.tar.gz. This entry and its partner entry 10.5281/zenodo.11397486 also share the intial coordinates used in the MD simulations of >800,000 docking poses of 4022 protein-ligand complexes (structureFiles_dockingPoses1.tar.gz in this entry and structureFiles_dockingPoses2.tar.gz in 10.5281/zenodo.11397486) and of the experimental ligand binding pose of 4549 complexes (structureFiles_experimentalStructures.tar.gz) as well as the corresponding ABFE estimates (absoluteBindingFreeEnergyEstimates.tar.gz). The MD simulations were run on the LUMI and MeluXina supercomputers while the implicit-solvent calculations were carried out on Galileo (Cineca).</p> <p>The README file describes the structure of the shared data in more detail and points out how to reproduce the MD trajectories and the subsequent implicit-solvent calculations yielding the free-energy estimates as well as how to use the data provided in this entry to train a machine-learning model predicting the ABFE of binding poses of protein-ligand complexes. The workflow scripts can be downloaded from GitHub (https://github.com/LigateProject/Pose-Selector-workflow). The MD simulations were run with GROMACS 2023.2 (https://manual.gromacs.org/2023.2/index.html), and the implicit-solvent calculations were carried out with gmx_MMPBSA 1.6.1 (https://valdes-tresanco-ms.github.io/gmx_MMPBSA/v1.6.1/).</p>
In situ conductometry for studying the homogenization of Al-Mg-Si alloys and predicting extrudate grain structure through machine learning
<p>This dataset includes the <em>in situ</em> impedance and time/temperature data from [1], grain structure data created by extrusion simulation coupled with physically-based microstructural simulation [2], and the predictions of the feed-forward neural network GRAINN-1/2 [1].</p> <p>[1] Österreicher, J. A., Zivanovic, D., Walenta, W., Maimone, S.,Hofbauer, M., Hovden, S., Tükör, Z., Arnoldt, A., Cerny, A. Kronsteiner, A., Antic, M., Zickler, G., Ehmeier, F., Mikulovic, M., Kunschert, G. (2024) . In situ conductometry for studying the homogenization of Al-Mg-Si alloys and predicting extrudate grain structure through machine learning. <em>Materials & Design</em>, 113070.</p> <p>[2] Hovden, S., Kronsteiner, J., Arnoldt, A., Horwatitsch, D., Kunschert, G., & Österreicher, J. A. (2024). Parameter study of extrusion simulation and grain structure prediction for 6xxx alloys with varied Fe content. <em>Materials Today Communications</em>, <em>38</em>, 108128.</p>
A theoretical framework for learning through structural plasticity
<p>This dataset contains the data and scripts used for the preprint:</p> <div> <div><span>Tiddia, G., Sergi, L., & Golosio, B.</span> <span>(</span><span>2023</span><span>)</span><span>.</span> <span>A theoretical framework for learning through structural plasticity</span><span>.</span> <span><em>ArXiv</em></span><span>. DOI:</span> <a href="https://doi.org/10.48550/arXiv.2307.11735" rel="nofollow">https://doi.org/10.48550/arXiv.2307.11735</a></div> </div> <p>The dataset contains a README file with all the instructions for compiling and running the model. </p> <p> </p>
Figure 3. Distribution of Q3 values-Secondary Structure Prediction of Protein using Resilient Back Propagation Learning Algorithm
<p>The estimated accuracy for the α- helices (QH), β- strands (QE), C-coil states (QC), and three<br> state together (Q3) for the system is shown in Figure 3.</p>
Figure 2. PAM250 matrix for the encoded sequence-Secondary Structure Prediction of Protein using Resilient Back Propagation Learning Algorithm
<p>The PAM matrix (Dayhoff et al., 1978) describes the probability that original amino acid<br> will be replaced by another amino acid over a defined evolutionary interval. The unit of<br> evolutionary divergence is defined as the interval in which 1% of the amino acids have been<br> changed between two sequences. The work uses PAM250, which assumes the occurrence of 250-<br> point mutations per 100 amino acids.<br> So, for the given the protein sequence GIVEQCCASVCSLYQLENYCN, A will be replaced<br> by 1 -3 0 1 -3 -1 0 5 -2 -3 -4 -2 -3 -5 0 1 0 -7 -5 -1 as shown in Figure 2.</p>
Figure 1: Snapshot of the CB396 dataset-Secondary Structure Prediction of Protein using Resilient Back Propagation Learning AlgorithmSecondary Structure Prediction of Protein using Resilient Back Propagation Learning Algorithm
<p>The dataset used for this work is CB396. This dataset contains 396 non-redundant sequences<br> derived from the 3Dee database created by Cuff and Barton (Cuff & Barton, 1999). It contains 396<br> proteins with their respective secondary structure as shown in Figure 1.</p>
ScienceDex guides
Understand access before you commit
These curated guides explain access requirements, typical timelines, costs, and reuse considerations for widely used research datasets.
Allen Brain Atlas
Allen Brain Atlas is an Allen Institute collection of brain map atlases, datasets, APIs, and analysis tools covering mouse, human, and non-human primate brain resources.
Annotated Behaviour and Observability Dataset (ABODe)
ABODe is a University of Edinburgh DataShare dataset for behavior classification in group-housed mice using home-cage video, identities, bounding boxes, ground-plate positions, and annotator labels.
DANDI Archive for NWB datasets
DANDI is a BRAIN Initiative archive for publishing and sharing neurophysiology data, including electrophysiology, optophysiology, and behavioral data packaged as NWB and related standards.
International Brain Laboratory public data
The International Brain Laboratory public data releases expose standardized mouse decision-making experiments, including Neuropixels recordings, widefield calcium imaging, behavior, and session metadata accessed through the ONE API.
OpenNeuro
OpenNeuro is a free, open platform for sharing neuroimaging datasets, with public search, dataset pages, and download paths for web, S3, DataLad, and the OpenNeuro CLI.