Skip to content

Latest commit

 

History

History
49 lines (26 loc) · 4.28 KB

README.md

File metadata and controls

49 lines (26 loc) · 4.28 KB

The completeness of metadata accompanying omics studies

Preprint Available

This project contains the links to the datasets and the figures that were used for our study : ["Improving the completeness of public metadata accompanying omics studies"]

Table of contents

Datasets

Extraction of metadata

We carefully examined a total of 3,125 samples across 29 studies. The original publications from journals were manually surveyed to gather information about the nine clinical phenotypes in question. The authors of these publications who own the data were contacted personally to obtain the complete data that was analyzed for that particular study. To extract metadata from the public repository, two Python scripts were used. These scripts are available here. The first script is used to get the files from the repository in the XML format. Further, the second script extracts the information from the XML file into a CSV file. These summary files from the repository can be found here and the data summarized from the original publication can be found here.

Description of metadata accompanying sepsis studies

There are four CSV files that were used to produce the results of the analysis.

  • sepsis_clinical_phenotypes.csv contains data regarding the number of times a particular clinical phenotype has been reported on each - the publication and the public repository. The total number of times the clinical phenotype was reported is a sum of the individual platforms. This is further expressed as a percentage.

  • sepsis_comparison.csv reports the number of clinical phenotypes that have been reported on each of the platforms for each of the cohorts. There were a total of nine clinical phenotypes that were considered. The total number of clinical phenotypes has been expressed as a percentage of the total (all nine clinical phenotypes being reported corresponds to 100%).

  • sepsis_completeness.csv was used to observe which of the cohorts were most and least complete. The number of clinical phenotypes reported for each cohort on the publication and the public repository was counted, summed and the total was expressed as a percentage.

  • sepsis_individual_phenonotypes contains data to calculate the most and least discrepancy between the individual phenotypes reported on both platforms.

Reproducing results

We have prepared Jupyter Notebooks that utilize the data described above to visualize and reproduce the results presented in our editorial.

Acknowledgements

We take this opportunity to specifically thank Jeremy Rotman for assisting to write the two Python scripts to extract metadata from NCBI GEO. We also thank Henry Fu for his help in the initial manual work of going through publications to accumulate data.

Contact

Please do not hesitate to contact us (serghei.mangul@gmail.com, anushkar@usc.edu) if you have any comments, suggestions, or clarification requests regarding the study or if you would like to contribute to the extended analysis involving more disease conditions.