Post Wed Apr 17, 2019 3:37 pm

Healthy control PBMC replicates-dataset release

Hi all,

In 2018, Stanford HIMC was involved with Dr. Andrew Pollard's lab at Oxford/OVG on a >400 sample vaccine study. This work was funded by the Gates Foundation. We also had panel advice and assistance from the Stanford labs of Dr. Catherine Blish (NK markers) and Dr. Eugene Butcher (chemokine receptors).

We barcoded various donor-timepoints as single-donor samples, and also included an aliquot of a previously-characterized healthy control donor PBMC sample, to use as a QC/QA reference. This was only surface-phenotyping; no ICS or phosphoflow stims.

The vaccine donor-timepoint data is still being analyzed by the Oxford group. However, after the Stanford HIMC used the healthy control donor files as QC/QA to ensure the overall quality of the data, there was no planned use for those healthy control files.


Dr. Pollard has graciously allowed the public release of these healthy control replicate files, and I have made them available at FlowRepository: http://flowrepository.org/id/FR-FCM-Z2YR

There are 59 files, with the following information (also available in the header and other information in the FR entry):
1. All cells are from the same draw of the same healthy control donor. The LRS chamber was received, PBMCs isolated by Ficoll, and cryopreserved as multiple vials stored in LN2 until use.
2. Each day, a single vial was thawed, and split appropriately to add 1 PBMC control sample to each donor-timepoint barcode set. For example, if 3 donors were thawed in a given day, then there would be three replicates of the PBMC control.
3. CD45 live-cell barcoding was used, in a 2-of-5 scheme.
4. All reagents are from the same commercial lot or in-house MAXPAR conjugation for the entire study.
5. There were 22 plates, run across approximately 4 months. There was a pipetting error in Plate 1, omitting one of the surface antibodies. Therefore, Plate 1 is not consistent with the rest of the study files, and so none of its healthy controls are in the above FR link.
6. That said, Plate 1's EQ bead data is fine. The way we normalized the data, we normalized Plate 1 using the Premessa R package (Finck et al-style normalization), then used the output Beads files as reference for all subsequent plates. If files had to be concatenated, they were concatenated on the Fluidigm software *prior to* normalization.
7. After normalization, the files were debarcoded using Premessa. The healthy control files posted are therefore Normalized and Debarcoded.
8. There were generally 2 Helios machines used to run the samples, but once or twice, a third instrument was used when one of the usual two instruments was down.
9. I have changed the naming of the files to better reflect the above information; these renamed files are exported from Mac FlowJo X.

** No cell events were removed in this process; therefore, anyone wanting to perform clustering analysis still *DOES* need to gate down to LiveIntactSinglets before proceeding.

- Filename examples:
"06Apr18_Helios1_Plate19_Sample1_HIMCctrl_cct.fcs"
This sample was run on 06Apr18, on Helios1. It was from Plate 19, and was concatenated prior to normalization from at least 2 original files.

"14Mar18_Helios2_Plate12_Sample3_HIMCctrl.fcs"
This sample was run on 14Mar18, on Helios2. It was from Plate 12, and was *not* concatenated.

Sample 1, Sample 2, and Sample 3 naming is arbitrary, and solely to distinguish the HIMC control files resulting from Debarcoding of the original Study donor Barcoded samples.


It is our hope that this dataset will be of general use to the CyTOF community. Particularly, we think it would be extremely helpful in efforts to build batch-effect correction algorithms (same-day replicates vs between-day replicates, etc)......something definitely needed as CyTOF studies get bigger and bigger.


Mike