Page 1 of 1

"Time" channel units

PostPosted: Thu Aug 08, 2019 9:26 pm
by dtelad11
I'm embarrassed to admit that I'm missing a rudimentary piece of CyTOF trivia: In the "Time" channel, what are the units? I.e, if 1,000 "time" passed, how much is that in seconds?

The FCS file is supposed to include a $TIMESTEP parameter but as far as I can see CyTOF FCS files don't have it. I searched through Fluidigm's documentation and the McGuire and Ashhurst book and I can find no mention of it.

Re: "Time" channel units

PostPosted: Thu Aug 08, 2019 9:34 pm
by BjornZ

Re: "Time" channel units

PostPosted: Thu Aug 08, 2019 9:39 pm
by dtelad11
Thanks Zach, should have used the Forum search function. Was this confirmed by Fluidigm at any point? Looking at a random FCS file, I have $BTIM of 16:47:32, $ETIM of 17:50:36, so 3,784 seconds. max(time) - min(time) = 3,289,977. 3,289,977 / 3,784 = 869.4442 ... which is pretty close to 1,000, but still more off than I would expected.

Re: "Time" channel units

PostPosted: Thu Aug 08, 2019 9:48 pm
by mleipold
Hi El-ad,

A couple questions:
1. Was your FCS file acquired normally, or was it reprocessed from the IMD? Reprocessing can occasionally give you time differences (https://www.fluidigm.com/binaries/conte ... igm%3Afile).
2. Was this max/min calculation done on a Fluidigm-normalized file? If so, were your beads OK (ie, no time interval excision due to low bead count, etc)?

I'm just trying to think of how Time might be affected......


Mike

Re: "Time" channel units

PostPosted: Thu Aug 08, 2019 10:19 pm
by BjornZ
Hi El-ad, Fluidigm hasn't confirmed it (to me at least), but it has provided sensible values for every dataset I've looked at, both short (minutes) and long (hours).

From my experience/observations, Fluidgm's $ETIM doesn't necessarily mean anything about the acquisition end time. Instead, it seems to be when the FCS Conversion step started (which can be a minute or more after acquisition ends). If you re-process an IMD file, it also seems to set $ETIM to the current time. $ETIM seems to consistently match the "FCS Conversion" history step timestamp in the embedded XML segment, bolstering that hypothesis. I've also seen files with $BTIM==$ETIM, so at this point I pretty much ignore those keywords.

Z

Re: "Time" channel units

PostPosted: Thu Aug 08, 2019 10:21 pm
by BjornZ
Mike alludes to a good point, $BTIM could start when the preview or acquisition delay starts, and I think the first event's TIME is always 0.

Re: "Time" channel units

PostPosted: Thu Aug 08, 2019 10:22 pm
by mleipold
Hi all,

I went back to the CyTOF1 manual, and the main instance of "time" in the parameters was in Settling time, which had units of Milliseconds (as Zach proposed).

-----
Going back to the original 2009 Bandura et al paper that describes the CyTOF1: https://pubs.acs.org/doi/10.1021/ac901049w
"A trigger delay (9000 ns) and the recording segment length (3072 ns) are set to allow digitization of the segment of the signal that corresponds to m/z = 125−215, with 1 ns sampling resolution.

In a typical cell analysis experiment, 2 min of raw data (∼28 GB) is recorded as a single continuous record, containing all 9 216 000 single spectral segments (3072 ns long each) generated during the 2 min.

A typical mass spectrum for m/z = 136−176 for the integral of 40 000 spectra (0.52 s acquisition time) for a sample of all (natural abundance) lanthanides at 20 pg/mL is shown in Figure 3.

For cell/particle analysis, the first stage of processing of the raw data record comprises integration of user-selected mass channels within each single spectrum, compressing the data by a factor of 3072/2N, where N is the number of selected mass channels (2 Bytes of data for each). When 30 mass channels are selected, the resulting compression factor is ∼50. The resulting files (typically <1 GB) contain “per spectrum” analog, pseudopulse, and dual data for the selected masses. Raw data for the beads shown in Figure 4 were collected for 120 s in high-resolution mode, and the data for each spectrum for m/z = 159, 165, and 169 were integrated within 16 ns-long integration windows, starting from ∼3% peak height. The resultant per-mass data for each of the first 64 000 consecutive spectra are shown in Figure 5.


Presumably those numbers relate to the IMD file, not the FCS file....I couldn't find any specific numbers for FCS writing. Since the FCS file is for the cell events, not the pushes like for the IMD, the "Time" has to be different.....


Mike

Re: "Time" channel units

PostPosted: Thu Aug 08, 2019 10:41 pm
by dtelad11
Thanks! I think this resolves the problem.