2024-Chen et al-preprint

[mleipold]

"Automated Cytometric Gating with Human-Level Performance Using Bivariate Segmentation"
Jiong Chen, Matei Ionita, Yanbo Feng, Yinfeng Lu, Patryk Orzechowski, Sumita Garai, Kenneth Hassinger, Jingxuan Bao, Junhao Wen, Duy Duong-Tran, Joost Wagenaar, Michelle L. McKeague, Mark M. Painter, Divij Mathew, Ajinkya Pattekar, Nuala J. Meyer, E. John Wherry, Allison R. Greenplate, Li Shen
bioRxiv, posted May 09, 2024
https://doi.org/10.1101/2024.05.06.592739

- will update when/if this is peer-reviewed and published

- analyzes 2 CyTOF datasets and 1 flow dataset
- COVID Vaccine Dataset (CyTOF) - appears to be new; no dataset accession given
- COVID vs Healthy Dataset (CyTOF) - appears to be new; no dataset accession given




34

[mleipold]

Now published:

Nat Commun. 2025, 16, 1576
https://doi.org/10.1038/s41467-025-56622-2

- "Data availability All datasets used in this study are publicly available and have been uploaded to the Pennsieve platform (https://doi.org/10.26275/864r-dv00), and all relevant source data are provided with the paper. In addition, the mass cytometry data used in this study are available in the FlowRepository database under accession code FR-FCM-Z8XU. The flow cytometry data used in this study are available in the FlowRepository database under accession code FR-FCM-Z8P9. This includes both mass cytometry and flow cytometry data, along with manual gating results from each individual annotator and the consensus gating outcomes. All data can be accessed via the provided link."

"We used two independent datasets consisting of single-cell protein expression data profiled by mass cytometry (Table 4). Both datasets were collected by the Institute for Immunology and Immune Health at the Perelman School of Medicine. For the first (“COVID Vaccine”) dataset, whole blood was obtained from 40 healthy subjects at four timepoints during the course of two-dose mRNA vaccination against COVID-19 and cryopreserved. (T1 = baseline, T2 = one week after the first dose, T3 = prior to the second dose, T4 = one week after the second dose). Only the baseline sample was used, in order to avoid information leaks across training. For the second (“Acute”) dataset, whole blood was obtained from 23 subjects, among whom 12 had acute COVID-19 symptoms and 11 were healthy donors. Samples collected for the COVID-19/healthy dataset were used fresh."


** The Pennsieve link (63 CSV files) and the FR-FCM-Z8XU (63 FCS files) both work. From Table 4 in the paper, this does work out to the correect number of FCS files (40 files in COVID Vaccine Dataset, 12+11 files in COVID Acute Dataset = 63 files).

However, it's incredibly unclear which files go with which study. Perhaps I missed it, but I can't find a file key in the accessions or in the Supplemental for the paper. It would be appreciated if any of the authors would clarify (or point out where I missed something).

For example, assuming that the "T" files are in COVID Vaccine study, the timepoints don't work out:
T1=35 files
T2=1 file
T3=1 file
T4=0 files
= 37 files (rather than 40 files)

There are 12 files with "HD" in the name (rather than "11 were healthy donors"?). And unclear what the the "ND_BM" and "ND_Zeus" files are either.