Dataset Search

A dataset of acoustic impulse responses for microphones worn on the body. Microphones were placed at 80 positions on the body of a human subject and a plastic mannequin. The impulse responses can be used to study the acoustic effects of the body and can be convolved with sound sources to simulate wearable audio devices and microphone arrays. The dataset also includes measurements with different articles of clothing covering some of the microphones and with microphones placed on different hats and accessories. The measurements were performed from 24 angles of arrival in an acoustically treated laboratory. Related Paper: Ryan M. Corey, Naoki Tsuda, and Andrew C. Singer. "Acoustic Impulse Responses for Wearable Audio Devices," IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP), Brighton, UK, May 2019. All impulse responses are sampled at 48 kHz and truncated to 500 ms. The impulse response data is provided in WAVE audio and MATLAB data file formats. The microphone locations are provided in tab-separated-value files for each experiment and are also depicted graphically in the documentation. The file wearable_mic_dataset_full.zip contains both WAVE- and MATLAB-format impulse responses. The file wearable_mic_dataset_matlab.zip contains only MATLAB-format impulse responses. The file wearable_mic_dataset_wave.zip contains only WAVE-format impulse responses.

This upload includes the 16S.B.ALL in 100-HF condition (referred to as 16S.B.ALL-100-HF) used in Experiment 3 of the WITCH paper (currently accepted in principle by the Journal of Computational Biology). 100-HF condition refers to making sequences fragmentary with an average length of 100 bp and a standard deviation of 60 bp. Additionally, we enforced that all fragmentary sequences to have lengths > 50 bp. Thus, the final average length of the fragments is slightly higher than 100 bp (~120 bp). In this case (i.e., 16S.B.ALL-100-HF), 1,000 sequences with lengths 25% around the median length are retained as "backbone sequences", while the remaining sequences are considered "query sequences" and made fragmentary using the "100-HF" procedure. Backbone sequences are aligned using MAGUS (or we extract their reference alignment). Then, the fragmentary versions of the query sequences are added back to the backbone alignment using either MAGUS+UPP or WITCH. More details of the tar.gz file are described in README.txt.

The synthetic networks in this dataset were generated using the RECCS protocol developed by Anne et al. (2024). Briefly, the RECCS process is as follows. An input network and clustering (by any algorithm) is used to pass input parameters to a stochastic block model (SBM) generator. The output is then modified to improve fit to the input real world clusters after which outlier nodes are added using one of three different options. See Anne et al. (2024): in press Complex Networks and Applications XIII (preprint : arXiv:2408.13647). The networks in this dataset were generated using either version 1 or version 2 of the RECCS protocol followed by outlier strategy S1. The input networks to the process were (i) the Curated Exosome Network (CEN), Wedell et al. (2021), (ii) cit_hepph (https://snap.stanford.edu/), (iii) cit_patents (https://snap.stanford.edu/), and (iv) wiki_topcats (https://snap.stanford.edu/). Input Networks: The CEN can be downloaded from the Illinois Data Bank: https://databank.illinois.edu/datasets/IDB-0908742 -> cen_pipeline.tar.gz -> S1_cen_cleaned.tsv The synthetic file naming system should be interpreted as follows: a_b_c.tsv.gz where a - name of inspirational network, e.g., cit_hepph b - the resolution value used when clustering a with the Leiden algorithm optimizing the Constant Potts Model, e.g., 0.01 c- the RECCS option used to approximate edge count and connectivity in the real world network, e.g., v1 Thus, cit_hepph_0.01_v1.tsv indicates that this network was modeled on the cit_hepph network and RECCSv1 was used to match edge count and connectivity to a Leiden-CPM 0.01 clustering of cit_hepph. For SBM generation, we used the graph_tool software (P. Peixoto, Tiago 2014. The graph-tool python library. figshare. Dataset. https://doi.org/10.6084/m9.figshare.1164194.v14) Additionally, this dataset contains synthetic networks generated for a replication experiment (repl_exp.tar.gz). The experiment aims to evaluate the consistency of RECCS-generated networks by producing multiple replicates under controlled conditions. These networks were generated using different configurations of RECCS, varying across two versions (v1 and v2), and applying the Connectivity Modifier (CM++, Ramavarapu et al. (2024)) pre-processing. Please note that the CM pipeline used for this experiment filters small clusters both before and after the CM treatment. Input Network : CEN Within repl_exp.tar.gz, the synthetic file naming system should be interpreted as follows: cen_<resolution><cm_status><reccs_version>sample<replicate_id>.tsv where: cen – Indicates the network was modeled on the Curated Exosome Network (CEN). resolution – The resolution parameter used in clustering the input network with Leiden-CPM (0.01). cm_status – Either cm (CM-treated input clustering) or no_cm (input clustering without CM treatment). reccs_version – The RECCS version used to generate the synthetic network (v1 or v2). replicate_id – The specific replicate (ranging from 0 to 2 for each configuration). For example: cen_0.01_cm_v1_sample_0.tsv – A synthetic network based on CEN with Leiden-CPM clustering at resolution 0.01, CM-treated input, and generated using RECCSv1 (first replicate). cen_0.01_no_cm_v2_sample_1.tsv – A synthetic network based on CEN with Leiden-CPM clustering at resolution 0.01, without CM treatment, and generated using RECCSv2 (second replicate). The ground truth clustering input to RECCS is contained in repl_exp_groundtruths.tar.gz.

Simulated sequences provide a way to evaluate multiple sequence alignment (MSA) methods where the ground truth is exactly known. However, the realism of such simulated conditions often comes under question compared to empirical datasets. In particular, simulated data often does not display heterogeneity in the sequence lengths, a common feature in biological datasets. In order to imitate sequence length heterogeneity, we here present a set of data that are evolved under a mixture model of indel lengths, where indels have an occasional chance of being promoted to long indels (emulating large insertion/deletion events, e.g., domain-level gain/loss). This dataset is otherwise (e.g., in GTR parameters) analogous to the 1000M condition as presented in the SATe paper (doi: 10.1126/science.1171243) but with 5000 sequences and simulated with INDELible (http://abacus.gene.ucl.ac.uk/software/indelible/). For more information, see README.txt. For the INDELible control files, see https://github.com/ThisBioLife/5000M-234-het.

Data sets to reproduce the results provided by the tutorial in paper "Data-Driven and Machine Learning Based Framework for Image-Guided Single-Cell Mass Spectrometry"

References - Li, Fu, Umberto Villa, Seonyeong Park, and Mark A. Anastasio. "3-D stochastic numerical breast phantoms for enabling virtual imaging trials of ultrasound computed tomography." IEEE Transactions on Ultrasonics, Ferroelectrics, and Frequency Control 69, no. 1 (2021): 135-146. DOI: 10.1109/TUFFC.2021.3112544 - Li, Fu; Villa, Umberto; Park, Seonyeong; Anastasio, Mark, 2021, "2D Acoustic Numerical Breast Phantoms and USCT Measurement Data", https://doi.org/10.7910/DVN/CUFVKE, Harvard Dataverse, V1 Overview - This dataset includes 1,089 two-dimensional slices extracted from 3D numerical breast phantoms (NBPs) for ultrasound computed tomography (USCT) studies. The anatomical structures of these NBPs were obtained using tools from the Virtual Imaging Clinical Trial for Regulatory Evaluation (VICTRE) project. The methods used to modify and extend the VICTRE NBPs for use in USCT studies are described in the publication cited above. - The NBPs in this dataset represent the following four ACR BI-RADS breast composition categories: > Type A - The breast is almost entirely fatty > Type B - There are scattered areas of fibroglandular density in the breast > Type C - The breast is heterogeneously dense > Type D - The breast is extremely dense - Each 2D slice is taken from a different 3D NBP, ensuring that no more than one slice comes from any single phantom. File Name Format - Each data file is stored as an HDF5 .mat file. The filenames follow this format: {type}{subject_id}.mat where{type} indicates the breast type (A, B, C, or D), and {subject_id} is a unique identifier assigned to each sample. For example, in the filename D510022534.mat, "D" represents the breast type, and "510022534" is the sample ID. File Contents - Each file contains the following variables: > "type": Breast type > "sos": Speed-of-sound map [mm/μs] > "den": Ambient density map [kg/mm³] > "att": Acoustic attenuation (power-law prefactor) map [dB/ MHzʸ mm] > "y": power-law exponent > "label": Tissue label map. Tissue types are denoted using the following labels: water (0), fat (1), skin (2), glandular tissue (29), ligament (88), lesion (200). - All spatial maps ("sos", "den", "att", and "label") have the same spatial dimensions of 2560 x 2560 pixels, with a pixel size of 0.1 mm x 0.1 mm. - "sos", "den", and "att" are float32 arrays, and "label" is an 8-bit unsigned integer array.

This repository contains the training dataset associated with the 2023 Grand Challenge on Deep Generative Modeling for Learning Medical Image Statistics (DGM-Image Challenge), hosted by the American Association of Physicists in Medicine. This dataset contains more than 100,000 8-bit images of size 512x512. These images emulate coronal slices from anthropomorphic breast phantoms adapted from the VICTRE toolchain [1], with assigned X-ray attenuation coefficients relevant for breast computed tomography. Also included are the labels indicating the breast type. The challenge has now concluded. More information about the challenge can be found here: <a href="https://www.aapm.org/GrandChallenge/DGM-Image/">https://www.aapm.org/GrandChallenge/DGM-Image/</a>. * New in V3: we added a CSV file containing the image breast type labels and example images (PNG).

This dataset contains the training results (model parameters, outputs), datasets for generalization testing, and 2-D implementation used in the article "Learned 1-D passive scalar advection to accelerate chemical transport modeling: a case study with GEOS-FP horizontal wind fields." The article will be submitted to Artificial Intelligence for Earth Systems. The datasets are saved as CSV for 1-D time-series data and *netCDF for 2-D time series dataset. The model parameters are saved in every training epoch tested in the study.

DNDC scripts and outputs that were generated as a part of the research publication 'Evaluation of DeNitrification DeComposition Model for Estimating Ammonia Fluxes from Chemical Fertilizer Application'.

This dataset contains sequencing data obtained from Illumina MiSeq device to prove the concept of the proposed 2DDNA framework. Please refer to README.txt for detailed description of each file.

This dataset contains the data for PASTA and UPP. PASTA data was used in the following articles: Mirarab, Siavash, Nam Nguyen, Sheng Guo, Li-San Wang, Junhyong Kim, and Tandy Warnow. “PASTA: Ultra-Large Multiple Sequence Alignment for Nucleotide and Amino-Acid Sequences.” Journal of Computational Biology 22, no. 5 (2015): 377–86. doi:10.1089/cmb.2014.0156. Mirarab, Siavash, Nam Nguyen, and Tandy Warnow. “PASTA: Ultra-Large Multiple Sequence Alignment.” Edited by Roded Sharan. Research in Computational Molecular Biology, 2014, 177–91. UPP data was used in: Nguyen, Nam-phuong D., Siavash Mirarab, Keerthana Kumar, and Tandy Warnow. “Ultra-Large Alignments Using Phylogeny-Aware Profiles.” Genome Biology 16, no. 1 (December 16, 2015): 124. doi:10.1186/s13059-015-0688-z.

This dataset provides the data for Nguyen, Nam-phuong, et al. "TIPP: taxonomic identification and phylogenetic profiling." Bioinformatics 30.24 (2014): 3548-3555.

Scripts and example data for AIDData (aiddata.org) processing in support of forthcoming Nakamura dissertation. This dataset includes two sets of scripts and example data files from an aiddata.org data dump. Fuller documentation about the functionality for these scripts is within the readme file. Additional background information and description of usage will be in the forthcoming Nakamura dissertation (link will be added when available). Data originally supplied by Nakamura. Python code and this readme file created by Wickes. Data included within this deposit are examples to demonstrate execution. Roughly, there are two python scripts in here: keyword_search.py, designed to assist in finding records matching specific keywords, and matching_tool.ipynb, designed to assist in detection of which records are and are not contained within a keyword results file and an aiddata project data file.

We are releasing the tracing dataset of four microservice benchmarks deployed on our dedicated Kubernetes cluster consisting of 15 heterogeneous nodes. The dataset is not sampled and is from selected types of requests in each benchmark, i.e., compose-posts in the social network application, compose-reviews in the media service application, book-rooms in the hotel reservation application, and reserve-tickets in the train ticket booking application. The four microservice applications come from [DeathStarBench](https://github.com/delimitrou/DeathStarBench) and [Train-Ticket](https://github.com/FudanSELab/train-ticket). The performance anomaly injector is from [FIRM](https://gitlab.engr.illinois.edu/DEPEND/firm.git). The dataset was preprocessed from the raw data generated in FIRM's tracing system. The dataset is separated by on which microservice component is the performance anomaly located (as the file name suggests). Each dataset is in CSV format and fields are separated by commas. Each line consists of the tracing ID and the duration (in 10^(-3) ms) of each component. Execution paths are specified in `execution_paths.txt` in each directory.

Sample data from one typical phantom test and one deidentified shunt patient test (shown in Fig. 8 of the MRM paper), with the corresponding analysis code for the Shunt-FENSI technique. For the MRM paper “Measuring CSF Shunt Flow with MRI Using Flow Enhancement of Signal Intensity (FENSI)”

HGT+ILS datasets from Davidson, R., Vachaspati, P., Mirarab, S., & Warnow, T. (2015). Phylogenomic species tree estimation in the presence of incomplete lineage sorting and horizontal gene transfer. BMC genomics, 16(10), 1-12. Contains model species trees, true and estimated gene trees, and simulated alignments.

Contains scattering data obtained for (TaSe4)2I at the Advanced Photon Source at Argonne National Laboratory. Beamline 6ID-D was used with a beam energy of 64.8 keV in a transmission geometry. Data was obtained at temperatures between 28 and 300 K. See the readme.txt file for more information.

This upload contains one additional set of datasets (RNASim10k, ten replicates) used in Experiment 2 of the EMMA paper (appeared in WABI 2023): Shen, Chengze, Baqiao Liu, Kelly P. Williams, and Tandy Warnow. "EMMA: A New Method for Computing Multiple Sequence Alignments given a Constraint Subset Alignment". The zipped file has the following structure: 10k |__R0 |__unaln.fas |__true.fas |__true.tre |__R1 ... # Alignment files: 1. `unaln.fas`: all unaligned sequences. 2. `true.fas`: the reference alignment of all sequences. 3. `true.tre`: the reference tree on all sequences. For other datasets that uniquely appeared in EMMA, please refer to the related dataset (which is linked below): Shen, Chengze; Liu, Baqiao; Williams, Kelly P.; Warnow, Tandy (2022): Datasets for EMMA: A New Method for Computing Multiple Sequence Alignments given a Constraint Subset Alignment. University of Illinois at Urbana-Champaign. https://doi.org/10.13012/B2IDB-2567453_V1

This dataset contains the data for 16S and 23S rRNA alignments including their reference trees. The original alignments are from the Gutell Lab CRW, currently located at https://crw-site.chemistry.gatech.edu/DAT/3C/Alignment/.

8-bit RGB realizations of a stochastic image model (SIM) of the **kinds** of things seen in fluorescence microscopy of biological samples. Note that no attempt was made to model a particular tissue, sample, or microscope. Distinct image features are seen in each color channel. The first public mention of these SIMs is in "Evaluation of Machine-generated Biomedical Images via A Tally-based Similarity Measure" by Frank Brooks and Rucha Deshpande. Manuscript on ArXiv and submitted for publication.

This dataset contains the data for SATe-I. SATe-I data was used in the following article: K. Liu, S. Raghavan, S. Nelesen, C. R. Linder, T. Warnow, "Rapid and Accurate Large-Scale Coestimation of Sequence Alignments and Phylogenetic Trees," Science, vol. 324, no. 5934, pp. 1561-1564, 19 June 2009.

This dataset consists of a citation graph. It was constructed by downloading and parsing the Works section of the Open Alex catalog of the global research system. Open Alex (see citation below) contains detailed information about scholarly research, including articles, authors, journals, institutions, and their relationships. The data were downloaded on 2024-07-15. The dataset comprises two compressed (.xz) files. 1) filename: openalexID_integer_id_hasDOI.parquet.xz. The tabular data within contains three columns: openalex_id, integer_id, and hasDOI. Each row represents a record with the following data types: • openalex_id: A unique identifier from the Open Alex catalog. • integer_id: An integer representing the new identifier (assigned by the authors) • hasDOI: An integer (0 or 1) indicating whether the record has a DOI (0 for no, 1 for yes). 2) filename: citation_table.tsv.xz This edgelist of citations has two columns (no header) of integer values that represent citing and cited integer_id, respectively. Summary Features • Total Nodes (Documents): 256,997,006 • Total Edges (citations): 2,148,871,058 • Documents with DOIs: 163,495,446 • Edges between documents with DOIs: 1,936,722,541 [corrected to 2,148,788,148 edges Nov 13, 2025] • Count of unique nodes in edgelist 111,453,719 [updated Nov 13, 2025] Note: Nov 13, 2025. An improved curation process will be applied to a future version of this dataset Note: Nov 13, 2025. The code used to generate these files can be found here: https://github.com/illinois-or-research-analytics/lorran_openalex/

Traces created using DeathStarBench (https://github.com/delimitrou/DeathStarBench) benchmark of microservice applications with injected failures on containers. Failures consist of disk/CPU/memory failures.