-
Notifications
You must be signed in to change notification settings - Fork 1
Data and metadata
In this demonstrator we make use of the data set published in Su et al. 2020.
The omics and phenotype data from Su et al. 2020 was prepared for upload to a FAIR Data Cube. Metadata has been added to describe the different data types. See the Tools wiki page for detailed information. The prepared data and metadata is stored at data/Su_2020_FAIR. The individual files are described in sections below.
data/Su_2020_FAIR/
├───i_investigation.txt
├───s_study.txt
├───a_assay_metabolomics.txt
├───a_assay_proteomics.txt
├───a_assay_transcriptomics.txt
│
├───metabolomics/
│ ├───metabolomics_Su_2020_feature-data.csv
│ └───metabolomics_Su_2020_feature-metadata.csv
│
├───proteomics/
│ ├───proteomics_Su_2020_feature-data.csv
│ └───proteomics_Su_2020_feature-metadata.csv
│
├───transcriptomics/
│ ├───transcriptomics_Su_2020_feature-data.csv
│ └───transcriptomics_Su_2020_feature-metadata.csv
│
├───phenotypes/
│ ├───individualIDs-vs-sampleIDs.ttl
│ ├───individualIDs-vs-sex.ttl
│ ├───individualIDs-vs-age.ttl
│ └───individualIDs-vs-phenotypic_features.ttl
The experimental data, as described in Su et al. 2020 is captured in the Investigation/Study/Assay (ISA) metadata framework (Sansone et al. 2012). A Jupyter Notebook that was used to create these ISA files can be found in tools/Su_2020_prepare_ISA_metadata. More information on the ISA framework can be found on the ISAtools website.
Metabolomics data has been extracted from Supplementary Table S1 from Su et al., 2020. The R notebook convert_data_metadata_chebi has been used to map metabolite features to ChEBi IDs, filter the matched entries based on a match score and an annotation score, and write two output files:
| Data type | File name | Description |
|---|---|---|
| Metabolomics data | metabolomics_Su_2020_feature-data.csv |
Contains 496 metabolites (after filtering) in rows and all original samples in columns |
| Metabolite feature annotations | metabolomics_Su_2020_feature-metadata.csv |
Contains feature annotations |
| feature.name | 1004596 | 1008097 | 1008631 |
|---|---|---|---|
| S-1-pyrroline-5-carboxylate | 0.78068456 | NA | 1.093995098 |
| spermidine | 1.914102595 | 0.722351202 | 0.410542328 |
| feature.name | database.ID | database | database.version |
|---|---|---|---|
| S-1-pyrroline-5-carboxylate | 17388 | https://www.ebi.ac.uk/chebi/ | ChEBI Release 211 |
| spermidine | 16610 | https://www.ebi.ac.uk/chebi/ | ChEBI Release 211 |
Transcriptomics data (i.e. scRNA-seq data from blood) has been downloaded from Array Express under the accession number: E-MTAB-9357. This original scRNA-seq data was CPM-normalized and then log-transformed (natural log1p) by the authors using methods from the SCANpy package. Further details on (generation and processing of) the transcriptomics data by the original authors are described in the methods sections about Single cell RNA-seq of the paper by: Su et al., 2020. Thereafter, for this TWOC demonstrator project, the single-cell RNA-seq data has been converted to pseudobulk RNA-seq data by taking the sum of all gene expression values for each individual gene, for each sample.
The ontologizing of features (i.e. gene symbols) was done by mapping to the GRCh38 human reference genome, only accepting perfect matches between the gene symbol ontology terms. Through this effort, we were able to assign 18,318 features of the total 24,966 (i.e. 6648 features remain unresolved, 26.6%). In the end, we wrote two output files:
| Data type | File name | Description |
|---|---|---|
| Transcriptomics data | transcriptomics_Su_2020_feature-data.csv |
Contains 18,318 genes (NB these are not filtered on expression level or on anything else) in rows and all T1 (*-BL) and control samples in columns |
| Transcriptomics feature annotations | transcriptomics_Su_2020_feature-metadata.csv |
Contains 18,318 gene annotations |
| feature.name | INCOV100-BL | INCOV101-BL | INCOV010-BL |
|---|---|---|---|
| A1BG | 451.835 | 883.444 | 4707.21 |
| A2M | 5.53521 | 0.00000 | 28.6070 |
| feature.name | database.ID | database | database.version |
|---|---|---|---|
| A1BG | NM_130786 | https://www.ncbi.nlm.nih.gov/nuccore/ | RefSeq Release 213 |
| A2M | NM_000014 | https://www.ncbi.nlm.nih.gov/nuccore/ | RefSeq Release 213 |
Proteomics data has been extracted from Supplementary Table S1 from Su et al., 2020. The proteomics feature names as reported by Su et al. were mapped to UniProt ID, through perfect string matching to the UniProtKB database (in case of multiple hits the top query result was selected, this happened in 113 of the 251 cases). Through this effort, we were able to confidently assign 251 of the total 464 features (i.e. 209 features remain unresolved, 45.0%; and 4 were removed because they were absent/undetected in all study samples). In the end, we wrote two output files:
| Data type | File name | Description |
|---|---|---|
| Proteomics data | proteomics_Su_2020_feature-data.csv |
Contains 251 proteins (NB these are not filtered on abundance level or on anything else) in rows and all samples in columns |
| Proteomics feature annotations | proteomics_Su_2020_feature-metadata.csv |
Contains 251 protein annotations |
| feature.name | 1010823 | 1027114 | 1029090 |
|---|---|---|---|
| eukaryotic translation initiation factor 4E binding protein 1 | 9.7486 | 7.0288 | 8.2784 |
| adenosine deaminase | 4.1470 | 4.0730 | 3.6326 |
| feature.name | database.ID | database | database.version |
|---|---|---|---|
| eukaryotic translation initiation factor 4E binding protein 1 | 4EBP1_HUMAN | https://www.uniprot.org/uniprotkb/ | UniProtKB/Swiss-Prot Release 2022_03 |
| adenosine deaminase | ADA_HUMAN | https://www.uniprot.org/uniprotkb/ | UniProtKB/Swiss-Prot Release 2022_03 |
The phenotypic data were provided by the original author Su et al. 2020. We have capture the data using the Phenopackets framework (https://github.com/phenopackets). The phenopackets framework includes a comprehensive data structure using ontology terms to categorise and connect different types of phenotypic data. This data structure were previously (partially) transformed into a RDF-schema by LUMC-BioSemantics (https://github.com/LUMC-BioSemantics/phenopackets-rdf-schema). Based on the RDF-schema from LUMC, we wrote yarrraml template files to convert the phenotypic data into RDF turtle format and stored in Fair Data Point. This process made use of cde-in-box infrastructure (https://github.com/ejp-rd-vp/cde-in-box), which were developed by teams who work the EJP-RD project (https://www.ejprarediseases.org/).