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.
14
datasets available to search
ShareScore release 0.9.0
Dataset results
14 results for “epitope prediction”
ESM-2 embeddings for TCR-Epitope Binding Affinity Prediction Task
<p>This is the accompanying dataset that was generated by the GitHub project: <a href="https://github.com/tonyreina/tdc-tcr-epitope-antibody-binding">https://github.com/tonyreina/tdc-tcr-epitope-antibody-binding</a>. In that repository I show how to create a machine learning models for predicting if a T-cell receptor (TCR) and protein epitope will bind to each other.</p> <p>A model that can predict how well a TCR bindings to an epitope can lead to more effective treatments that use immunotherapy. For example, in anti-cancer therapies it is important for the T-cell receptor to bind to the protein marker in the cancer cell so that the T-cell (actually the T-cell's friends in the immune system) can kill the cancer cell.</p> <div> <div>[HuggingFace](https://huggingface.co/facebook/esm2_t36_3B_UR50D) provides a "one-stop shop" to train and deploy AI models. In this case, we use Facebook's open-source [Evolutionary Scale Model (ESM-2)](https://github.com/facebookresearch/esm). These embeddings turn the protein sequences into a vector of numbers that the computer can use in a mathematical model.</div> <div> </div> To load them into Python use the Pandas library:</div> <pre><code>import pandas as pd train_data = pd.read_pickle("train_data.pkl") validation_data = pd.read_pickle("validation_data.pkl") test_data = pd.read_pickle("test_data.pkl")</code></pre> <p>The <strong>epitope_aa</strong> and the <strong>tcr_full</strong> columns are the protein (peptide) sequences for the epitope and the T-cell receptor, respectively. The letters correspond to the <a href="https://en.wikipedia.org/wiki/DNA_and_RNA_codon_tables">standard amino acid codes</a>.</p> <p>The <strong>epitope_smi</strong> column is the <a href="https://en.wikipedia.org/wiki/Simplified_molecular-input_line-entry_system">SMILES</a> notation for the chemical structure of the epitope. We won't use this information. Instead, the ESM-1b embedder should be sufficient for the input to our binary classification model.</p> <p>The <strong>tcr</strong> column is the CDR3 hyperloop. It's the part of the TCR that actually binds (assuming it binds) to the epitope.</p> <p>The <strong>label</strong> column is whether the two proteins bind. 0 = No. 1 = Yes.</p> <p>The <strong>tcr_vector</strong> and <strong>epitope_vector</strong> columns are the bio-embeddings of the TCR and epitope sequences generated by the Facebook ESM-1b model. These two vectors can be used to create a machine learning model that predicts whether the combination will produce a successful protein binding.</p> <p>From the TDC website:</p> <blockquote> <p>T-cells are an integral part of the adaptive immune system, whose survival, proliferation, activation and function are all governed by the interaction of their T-cell receptor (TCR) with immunogenic peptides (epitopes). A large repertoire of T-cell receptors with different specificity is needed to provide protection against a wide range of pathogens. This new task aims to predict the binding affinity given a pair of TCR sequence and epitope sequence.</p> <p>Weber et al.</p> </blockquote> <p>Dataset Description: The dataset is from Weber et al. who assemble a large and diverse data from the VDJ database and ImmuneCODE project. It uses human TCR-beta chain sequences. Since this dataset is highly imbalanced, the authors exclude epitopes with less than 15 associated TCR sequences and downsample to a limit of 400 TCRs per epitope. The dataset contains amino acid sequences either for the entire TCR or only for the hypervariable CDR3 loop. Epitopes are available as amino acid sequences. Since Weber et al. proposed to represent the peptides as SMILES strings (which reformulates the problem to protein-ligand binding prediction) the SMILES strings of the epitopes are also included. 50% negative samples were generated by shuffling the pairs, i.e. associating TCR sequences with epitopes they have not been shown to bind.</p> <blockquote> <p>Task Description: Binary classification. Given the epitope (a peptide, either represented as amino acid sequence or as SMILES) and a T-cell receptor (amino acid sequence, either of the full protein complex or only of the hypervariable CDR3 loop), predict whether the epitope binds to the TCR.</p> <p>Dataset Statistics: 47,182 TCR-Epitope pairs between 192 epitopes and 23,139 TCRs.</p> <p>References:</p> </blockquote> <ol> <li>Weber, Anna, Jannis Born, and María Rodriguez Martínez. “TITAN: T-cell receptor specificity prediction with bimodal attention networks.” Bioinformatics 37.Supplement_1 (2021): i237-i244.</li> <li>Bagaev, Dmitry V., et al. “VDJdb in 2019: database extension, new analysis infrastructure and a T-cell receptor motif compendium.” Nucleic Acids Research 48.D1 (2020): D1057-D1062.</li> <li>Dines, Jennifer N., et al. “The immunerace study: A prospective multicohort study of immune response action to covid-19 events with the immunecode™ open access database.” medRxiv (2020).</li> </ol> <blockquote> <p>Dataset License: CC BY 4.0.</p> <p>Contributed by: Anna Weber and Jannis Born.</p> </blockquote> <p> </p> <div>The Facebook ESM-2 model has the MIT license and was published in:</div> <div> </div> <div>* Zeming Lin et al, Evolutionary-scale prediction of atomic-level protein structure with a language model, Science (2023). DOI: 10.1126/science.ade2574 https://www.science.org/doi/10.1126/science.ade2574</div> <div> </div> <div>HuggingFace has several versions of the trained model.</div> <div> </div> <div> <table> <tbody> <tr> <td>Checkpoint name</td> <td>Number of layers</td> <td>Number of parameters</td> </tr> <tr> <td>esm2_t48_15B_UR50D</td> <td>48</td> <td>15B</td> </tr> <tr> <td>esm2_t36_3B_UR50D</td> <td>36</td> <td>3B</td> </tr> <tr> <td>esm2_t33_650M_UR50D</td> <td>33</td> <td>650M</td> </tr> <tr> <td>esm2_t30_150M_UR50D</td> <td>30</td> <td>150M</td> </tr> <tr> <td>esm2_t12_35M_UR50D</td> <td>12</td> <td>35M</td> </tr> <tr> <td>esm2_t6_8M_UR50D</td> <td>6</td> <td>8M</td> </tr> </tbody> </table> </div>
Dataset: Predicting T Cell Receptor Functionality against Mutant Epitopes
<p>This upload contains the Datasets and Supplementary Material for "Predicting T Cell Receptor Functionality against Mutant Epitopes" (DOI: https://doi.org/10.1101/2023.05.10.540189) by Felix Drost, Emilio Dorigatti, Adrian Straub, Philipp Hilgendorf, Karolin I. Wagner, Kersten Heyer, Marta López Montes, Bernd Bischl, Dirk H. Busch, Kilian Schober, and Benjamin Schubert.</p>
In silico Antibody-Peptide Epitope prediction for Personalized cancer therapy publication data
<p>In silico Antibody-Peptide Epitope prediction for Personalized cancer therapy publication data</p>
Data for "Current challenges for unseen-epitope TCR interaction prediction and a new perspective derived from image classification" (ImRex)
<p>Repository containing the different experiments described in the manuscript titled: "Current challenges for epitope-agnostic TCR interaction prediction and a new perspective derived from image classification".</p> <p>Publication DOI: TBA</p> <p>Originally appeared as a preprint on bioRxiv: <a href="https://doi.org/10.1101/2019.12.18.880146">https://doi.org/10.1101/2019.12.18.880146</a>.</p> <p>Contains:</p> <ul> <li>Trained model files (.h5)</li> <li>Associated train and validation datasets for each model.</li> <li>Learning curves and evaluation metrics.</li> <li>Log files with training and data arguments (full training scripts are available in GitHub repository).</li> <li>Comparisons between different models.</li> <li>Complete raw and processed datasets (also available in the associated GitHub repository).</li> </ul> <p><strong>Please refer to the associated GitHub repository (<a href="https://github.com/pmoris/ImRex">https://github.com/pmoris/ImRex</a>) for more information on the directory structure and contents, as well as the scripts that generated these output files.</strong></p> <p><strong>Contents:</strong></p> <ul> <li> <p><code>data.zip</code>: Contains raw and preprocessed datasets. READMEs in subdirectory describe the data sources and preprocessing steps. Please refer to the associated GitHub repository for the specific scripts that generated these files. Note that the full training and test sets (i.e. containing both positive and negative examples) are stored separately for each model/CV iteration in the <code>models</code> archives.</p> </li> <li> <p><code>models-main.zip</code>: contains the trained models and evaluation metrics for the main different experiments described in the bash and pbs scripts in <code>./src/scripts/hpc_scripts</code>. Log files for the experiments outlined here can be found in <code>./src/scripts/hpc_scripts</code>.</p> </li> <li> <p><code>models-full.zip</code>: contains models that were trained on the complete VDJdb dataset without cross-validation, filtered on human TRB data, no 10x data and restricted to 10-20 (CDR3) or 8-11 (epitope) amino acid residues, with negatives that were generated by shuffling (i.e. sampling an negative epitope for each positive CDR3 sequence). One set of models uses downsampling to reduce the most abundant epitopes down to 400 pairs each, the other one does not use any downsampling. These models were also used for evaluating on the external Adaptive dataset, as outlined in <code>./src/scripts/evaluate/evaluate_adaptive.sh</code>, and the TRA subset of sequences (<code>./src/scripts/evaluate/evaluate_tra.sh</code>).</p> </li> <li> <p><code>models-decoyfit.zip</code>: contains models that were trained on true data, but evaluated on data where epitopes were replaced by decoys.</p> </li> <li> <p><code>models-padded-epitoperatio.zip</code>: contains a quick test of trained models (padded/interaction map) that use a different type of negative shuffling, see docstrings in <code>./src/processing/negative_sampler.py</code> for more info.</p> </li> <li> <p><code>models-repeat-local.zip</code>: contains a number of repeated runs from <code>models-main</code>, used to estimate variability in model performance for multiple identical runs.</p> </li> <li> <p><code>comparisons.zip</code>: contains comparison directories, each consisting of two or more model output directories, that contrast the performance metrics of the models. These outputs were generated by using the <code>./src/scripts/evaluate/visualize.py</code> script, or by using the oneliners in <code>./src/scripts/evaluate/visualise.sh</code>, which can operate on the entire comparisons directory at once.</p> </li> </ul> <p><strong>Note that any file paths described here are in reference to the associated GitHub repository (<a href="https://github.com/pmoris/ImRex">https://github.com/pmoris/ImRex</a>).</strong></p> <p><strong>Overview of different experiments:</strong></p> <ul> <li>Two main architectures were compared: the interaction map (or <code>padded</code>) CNN and a dual input CNN based on NetTCR (<code>nettcr</code>).</li> <li>Two different cross-validation strategies were used: a 5x repeated 5-fold CV (<code>repeated5fold</code>) and an epitope-grouped CV (<code>epitope_grouped</code>).</li> <li>The different dataset subsets are labelled as follows. Check the Makefile's <code>preprocess-vdjdb-aug-2019</code> command (and the underlying script <code>./src/scripts/preprocessing/preprocess_vdjdb.py</code>) for a more thorough overview of the different filtering options. <ul> <li><code>mhci</code>: only MHCI class presented epitopes.</li> <li><code>trb</code>: only TRB CDR3 sequences.</li> <li><code>tra</code>: only TRA CDR3 sequences.</li> <li><code>tratrb</code>: both types of CDR3 sequences.</li> <li><code>down</code>: moderate downsampling of most abundant epitopes to 1000 pairs.</li> <li><code>down400</code>: strong downsampling of most abundant epitopes to 400 pairs.</li> <li><code>decoy</code>: decoy epitope data.</li> <li><code>reg001</code>: regularization factor 0.01 (only for padded/interaction type models, fixed value)</li> </ul> </li> <li>Two different methods of generating negative TCR-epitope pairs were used: shuffling of positive pairs, i.e. sampling a single epitope from the positive pairs for each CDR3 sequence (<code>shuffle</code>), and sampling CDR3s from a reference repertoire (<code>negref</code>).</li> <li>The batch size is labelled as <code>b32</code> = a batch size of 32.</li> <li>The learning rate was always 0.0001 (<code>lre4</code>) or 0.001 (<code>lre3</code>).</li> </ul> <p> </p>
Identification and validation of 174 COVID‑19 vaccine candidate epitopes reveals low performance of common epitope prediction tools
<p>Datasets used in "Identification and validation of 174 COVID‑19 vaccine candidate epitopes reveals low performance of common epitope prediction tools"</p> <p>sars-cov-2_predicted.csv</p> <p>SARS-CoV-2 peptides predicted to be binders to 11 HLA alleles by the NetMHC suite tools, these selected peptides have been measured in a stability assay revealing SARS-CoV-2 epitopes (assay results can be found at: https://www.immunitrack.com/free-coronavirus-report-for-download/). </p> <p>tools_predictions_binders.xlsx</p> <p>File containing all predictions for tools tested in the benchmark (note that not all tools use same measure). List of binders (epitopes) with IEDB links and T cell studies overlap.</p>
Mapping Synthetic Binding Proteins Epitopes on Diverse Protein Targets by Protein Structure Prediction and Protein-Protein Docking
<p>The predicted 3D structures of 145 SBPs and the 96 models of SBPs in complex with protein targets.</p>
Data from: A general approach for predicting protein epitopes targeted by antibody repertoires using whole proteomes
Open the record for dataset details and reuse information.
Autoantibody Epitope Spreading in the Pre-Clinical Phase Predicts Progression to Rheumatoid Arthritis [ANALYTE: Cytokine or chemokine]
GEO Series GSE32019. Homo sapiens. 556 samples. Type: Protein profiling by protein array.
Autoantibody Epitope Spreading in the Pre-Clinical Phase Predicts Progression to Rheumatoid Arthritis [ANALYTE: ANTIGEN]
GEO Series GSE32016. Homo sapiens. 559 samples. Type: Protein profiling by protein array.
Machine learning predictions of MHC-II specificities reveal alternative binding mode of class II epitopes
GEO Series GSE205588. Homo sapiens. 8 samples. Type: Other.
Improved prediction of endogenous HLA-associated epitopes based on mono-allelic mass spectrometry profiling
GEO Series GSE93315. Homo sapiens. 4 samples. Type: Expression profiling by high throughput sequencing.
A large peptidome dataset improves HLA class I epitope prediction across most of the human population
GEO Series GSE131267. Homo sapiens. 8 samples. Type: Expression profiling by high throughput sequencing.
Autoantibody Epitope Spreading in the Pre-Clinical Phase Predicts Progression to Rheumatoid Arthritis
GEO Series GSE32021. Homo sapiens. 1115 samples. Type: Protein profiling by protein array.
Improved predictions of immunogenicity reveal SARS-Cov-2 CD8 T-cell epitopes
GEO Series GSE201212. Homo sapiens. 20 samples. Type: Other.
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.