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.
261
datasets available to search
ShareScore release 0.9.0
Dataset results
261 results for “protein prediction”
Protein structure data for "AI-predicted protein deformation encodes energy landscape perturbation"
<p>AF2-predicted protein structures of WT and mutant proteins that have corresponding ddG measurements in the ThermoMutDB database of protein mutant stability measurements. PDB structures are compressed using <a href="https://github.com/steineggerlab/foldcomp/">FoldComp</a>, and saved in "structures.zip".</p> <p>Summary of the final dataset and results can be found in "results_summary.pkl".</p> <p>Code used to plot figures can be found in "code4figs.zip".</p>
Predicting the pro-longevity or anti-longevity effect of model organism genes with enhanced Gaussian noise augmentation-based contrastive learning on protein-protein interaction networks
<p>The datasets used to evaluate Enhanced Gaussian noise augmentation-based contrastive learning (EGsCL) against predicting the pro-longevity or anti-longevity effect of model organism gene. This repo also includes the pretrained encoders that obtained the best predictive performance for each organism (see Table 2).</p>
Predicted protein-protein interactions of the human tyrosine kinase Lck by AF2Complex
<p>Structural models and Supplementary data described in the publication:</p> <p>Predicting protein interactions of the kinase Lck critical to T cell modulation</p> <p>Structure (Cell Press), 2024</p> <p>Reference Authors: Mu Gao and Jeffrey Skolnick</p> <p><br>screening_results -- Virtual PPI screening results of the protein kinase Lck (SH3-SH2 domains as the bait).<br>predicted structural models.zip -- Compressed structural models described in the publication.<br>af2c_input_features -- Input features used to predict protein complex models with AF2Complex. </p>
Exploring zero-shot structure-based protein fitness prediction
<p>This repository contains data used in Exploring zero-shot structure-based protein fitness<br>prediction.</p> <p>Directions to use this data can be found on <a href="https://github.com/gitter-lab/benchmarking-structure-based-models">our GitHub repository</a>.</p> <ol> <li><code>experimental_struct_artifacts</code> contains the experimentally determined structures for ProteinGym assays used in our analysis along with the reference file needed to generate ESM inverse folding predictions for these structures in ProteinGym.</li> <li><code>results</code> contains the prediction results obtained by running SSEmb on the 216 ProteinGym assays being considered in this study.</li> <li><code>test.tar.gz</code> contains all the structures from ProteinGym as well as MSAs generated using mmseqs2. To use this directory: <ul> <li>Setup SSEmb as directed in its <a href="https://github.com/KULL-Centre/_2023_Blaabjerg_SSEmb">repository</a></li> <li>Download this file and extract it in the data folder.</li> </ul> </li> </ol> <p> </p> <p> </p>
Assessing interaction recovery of predicted protein-ligand poses
<p>We provide the following data:</p> <ol> <li>PDB files containing protein structures from the PoseBusters dataset prepared with OpenEye's SPRUCE protein preparation software</li> <li>SDF files containing the corresponding PoseBusters ligand in its crystal pose</li> </ol> <p>This data is provided for the 256 PoseBuster targets used in our paper "Assessing interaction recovery of predicted protein-ligand poses" [1] with associated code at https://github.com/Exscientia/plif_validity.</p> <p> </p> <h3>References</h3> <p>[1] Errington D, Schneider C, Bouysset C, Dreyer FA, Assessing interaction recovery of predicted protein-ligand poses, arXiv; 2024. Available from: https://arxiv.org/abs/2409.20227 </p>
Predicted structures of the periplasmic adaptor protein, CmeA
<p>Predicted structures for CmeA, sequence alignments, and a fully assembled model for CmeABC (Chapter 6 of Kahlan Newman's Doctoral Thesis). </p>
Disorder predictions for all reviewed UniProt proteins with "membrane" subcellular localization
<h1>IDRs in membrane proteins</h1> <h3>About</h3> <p>The writeup below walks through how to answer the question "what fraction of membrane proteins have IDRs".</p> <p>The tl/dr: on average <strong>55% of membrane proteins [as defined here] have IDRs</strong>.</p> <h3>Data</h3> <p>This directory (WITH the associated sequences), which are not included on GitHub, are deposited on Zenodo.</p> <h2>Steps</h2> <h4>Install software</h4> <p>Install metapredict</p> <div> <pre><code>pip install metapredict</code></pre> </div> <p>Done.</p> <h4>Find the data</h4> <p>Using the "Advanced Search", I downloaded all reviewed proteins on <a href="https://uniprot.org/">UniProt</a> tagged as having a "membrane" subcellular localization.</p> <p>There are, of course, much more detailed ways one could get these data (e.g., requiring transmembrane helices, integral membrane protein filtering etc, etc), but I figured for purposes here, I'd just do this...</p> <h4>Download the data</h4> <p>From this, I downloaded all the FASTA sequences. This triggers the following file to be downloaded:</p> <div> <pre><code>uniprotkb_cc_scl_term_SL_0162_AND_revie_2024_11_09.fasta</code></pre> </div> <h4>Preprocessing the data</h4> <p>Next I wanted to fix any sequences with non-standard amino acids by converting those residues to "standard" amino acids. I did this using <code>pfasta</code>, a commandline tool that's part of <a href="https://protfasta.readthedocs.io/en/latest/">protfasta</a> which comes with metapredict.</p> <p>Did this simply with</p> <div> <pre><code>pfasta --invalid-sequence convert-all -o membrane_proteins.fasta uniprotkb_cc_scl_term_SL_0162_AND_revie_2024_11_08.fasta</code></pre> </div> <ul> <li>The <code>--invalid-sequence</code> flag tells pfasta what to with an invalid sequence, and <code>-o</code> tells it what the ouput file should be called.</li> </ul> <p>Running this (~5 seconds) generates the following output:</p> <div> <pre><code>........................ pfasta version 0.1.13+0.gc1786bd.dirty Please report bugs to: https://github.com/holehouse-lab/protfasta ........................ [INFO]: Reading in the file uniprotkb_cc_scl_term_SL_0162_AND_revie_2024_11_08.fasta [INFO]: Read in file with 1004330 lines [INFO]: Parsed file to recover 117691 sequences [INFO]: Converted 697 sequences to valid sequences [INFO]: Writing new sequence file [membrane_proteins.fasta]...</code></pre> </div> <h4>Predicting all IDRs</h4> <p>Run one of metapredict's command-line utilities to predict all IDRs in these proteins.</p> <div> <pre><code>metapredict-predict-idrs membrane_proteins.fasta -o membrane_protein_idrs.fasta</code></pre> </div> <p>This takes ~20 seconds on my desktop with an A4500 GPU or ~5 min with my Macbook Pro. We're then left with a file called <code>membrane_protein_idrs.fasta</code>. This is a FASTA file that the header format:</p> <div> <pre><code>><HEADER FOR ORIGINAL PROTEIN> IDR_START=<start position> IDR_END=<end_position></code></pre> </div> <p>e.g.</p> <div> <pre><code>>sp|B4R1D8|HH_DROSI Protein hedgehog OS=Drosophila simulans OX=7240 GN=hh PE=3 SV=1 IDR_START=0 IDR_END=94 MDNHSSVPWASAASVTCLSLDAKCHSSSSSCSSKSTASSISASPETQTMRHIAHTQRCLS RLTSLVALLLIVLPMMFSPAHSCGPGRGLGRHRA</code></pre> </div> <h3>Hacky way to get number of proteins with IDRs</h3> <p>Get the number of unique proteins in the main file:</p> <div> <pre><code>cat membrane_proteins.fasta | grep ">" | awk -F "|" {' print $2 '} | sort | uniq |wc</code></pre> </div> <p>Which returns</p> <div> <pre><code>117691 117691 829609</code></pre> </div> <p>So we have 117,691 unique proteins (NB: we could also just have gotten this without the <code>awk -F "|" {' print $2 '} | sort | uniq</code> part BUT we do need that for the number of unique proteins with IDRs so good to be consistent.</p> <div> <pre><code>cat membrane_protein_idrs.fasta | grep ">" | awk -F "|" {' print $2 '} | sort | uniq |wc</code></pre> </div> <p>Which returns</p> <div> <pre><code>64971 64971 458829 </code></pre> </div> <p>So on average <strong>55% of membrane proteins [as defined here] have IDRs</strong>.</p> <h3>Bonus!</h3> <p>In addition we can predict all disorder scores rather than discrete IDRs using</p> <div> <pre><code>metapredict-predict-disorder membrane_proteins.fasta -o membrane_protein_disorder_scores.csv</code></pre> </div> <p>This takes ~5 min on my Macbook Pro. The data for these are included here for completeness/convenience.</p>
The impact of the cross-docked poses on the performance of machine learning classifier for protein-ligand binding pose prediction
<p>Datasets, features, and some representative scripts utilized in the paper "The impact of the cross-docked poses on the performance of machine learning classifier for protein-ligand binding pose prediction".</p>
The dataset used in "A max-margin model for predicting residue-base contacts in protein-RNA interactions"
<p>The dataset used in "A max-margin model for predicting residue-base contacts in protein-RNA interactions"</p>
Structure prediction from SARS-CoV-2 accessory proteins ORF-6
<p>Structure prediction made with Collabfold for SARS-CoV-2 accessory protein ORF-6.</p> <p>The archive contains both the structure and</p>
Structure prediction from SARS-CoV-2 accessory proteins ORF-7B
<p>Structure prediction made with Collabfold for SARS-CoV-2 accessory protein ORF-7B.</p> <p>The archive contains both the structure and the logs from the prediction.</p>
Predicted protein functions for 2031 Saccharomyces cerevisiae genome assemblies
<p>Predicted protein functions for 2031 Saccharomyces cerevisiae genome assemblies</p>
DOX_BDW: Incorporating Solvation and Desolvation Effects of Cavity Water into Nonfitting Protein–Ligand Binding Affinity Prediction
<p><strong>structures.zip:</strong> including the coordinates of all optimized proteinligand complex structure obtained by DOX_BDW calculation. (compressed PDB file). These pdb files could also be used as input for the binding energy calculation,as illustrated in SI section 8. </p> <p><strong>mdinput.zip:</strong> Including the input files,parameter files, topology files needed to run MD simulation for water mapping, as illustrated in SI section 8. Note that all of the parameter files and topology files would be automatically generated using the RUNMD program we uploaded with the example file. </p> <p><strong>example.zip:</strong> The programs and input files needed to run an example, as illustrated in SI section 9. And all the output files except MD trajectories are in there,too.</p>
Structure-based self-supervised learning enables ultrafast prediction of stability changes upon mutation at the protein universe scale
<p>Pythia computed all single mutations of <em>E.coli</em> proteome, high quality high quality of Swiss-Prot structures and thermophilic proteins used in analysis.</p>
Dataset associated with the study titled "Systematic discovery of regulatory motifs associated with human insulator sites" It includes data used for predicting insulator-associated DNA-binding proteins.
<p><strong>This dataset was used in the study on the prediction of insulator-associated DNA-binding proteins.</strong></p> <p>The three files included here are intended for use with the code available in the following GitHub repository:<br><a href="https://github.com/reposit2/insulator" target="_new" rel="noopener">https://github.com/reposit2/insulator</a></p> <p>To use the data, extract all three files and place them in the same directory as the code files.</p>
A3D database: structure-based predictions of protein aggregation for the human proteome
<p>A3D database: structure-based predictions of protein aggregation for the human proteome</p>
Monte Carlo Arithmetic Instrumented DeepGOPlus Protein Function Predictions
<p>This dataset contains the perturbed protein function predictions by the DeepGOPlus model excluding the Diamond tool component. The model was perturbed with Verrou, an implementation of Monte Carlo Arithmetic (MCA), a stochastic arithmetic technique that injects noise into a program that simulates changes in a user's execution environment. The folders contain pkl files that can be read with the Pandas python library to load up dataframes containing the predictions and original values. Each file is one MCA sample run across the entire DeepGOPlus test set.</p> <p>The folder "Verrou_All" contains predictions where the entirety of the model was instrumented with MCA.</p> <p>The folder "Verrou_TF" contains predictions where only the Tensorflow library was instrumented with MCA.</p> <p>The folder "Fuzzy_Python" contains predictions where only the Python interpreter was instrumented with MCA.</p> <p>The folder "VPREC_Outbound_Mode" contains predictions where the virtual precision of the floating point operations was reduced witht the VPREC precision simulator tool in outbound mode.</p> <p>The folder "VPREC_Inbound_Mode" contains predictions where the virtual precision of the floating point operations was reduced witht the VPREC precision simulator tool in inbound mode.</p> <p>More information can be found by consulting <a href="https://arxiv.org/abs/2212.06361">this paper</a> or <a href="https://github.com/big-data-lab-team/deepgoplus-stability">this Github repository</a>.</p>
Identification of Predictive Factors for the Response to Anti-Programmed Cell Death Protein 1 (PD1) Immunotherapy in Head and Neck Squamous Cell Carcinoma
ClinicalTrials.gov study NCT05328024. IPD Sharing: Not stated. Countries: 1. Publications: 1.
The Peripheral(-Muscle) Oxygenation and Perfusion Score as a New Non-invasive Tool to Predict Elevations in C-reactive Protein Levels in Neonates
ClinicalTrials.gov study NCT07109856. IPD Sharing: NO. Countries: 1. Publications: 18.
A Pilot Study of a Protein Profile Test in Ovarian Cancer Patients in Remission to See if Protein Changes Can Predict Relapse (be Predictive of Cancer Relapse)
ClinicalTrials.gov study NCT00001938. IPD Sharing: Not stated. Countries: 1. Publications: 3.
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.