FlowClean#

This module gates events from time slices whose density is low or whose events’ fluorescence intensity varies substantially from other slices. This is often due to a bubble or transient clog in the flow cell.

The operation assesses whether a tube is “clean” by looking for changes in fluorescence over time, both fast changes (discontinuities) and slow changes (drift). If the tube is clean, only low-density slices are removed. If the tube is not clean, then a cleaning is attempted, removing slices that are substantially statistically different than the majority. Cleanliness is then assessed again.

This operation is applied to every tube independently – that is, to every subset of events with a unique combination of experimental metadata.

Name: The name of the gate you can subsequently use to exclude “unclean” events.

Time channel: The channel that represents time – often HDR-T or similar.

Channels: Which fluorescence channel or channels are analyzed for variation, and how should they be scaled before cleaning, and how they should be scaled.

Important

This algorithm works much better when fluorescence channels are using a scale other than linear!

Segment Size (default = 500): The number of events in each bin in the analysis.

Density Cutoff (default = 0.05): The minimum density CDF to keep.

Max Discontinuity (default = 0.1): The critical “continuity” – determines how “different” an adjacent segment must be to be for a tube to be flagged as suspicious.

Max Drift : Float (default = 0.15): The maximum any individual channel can drift before being flagged as needing cleaning.

Max Mean Drift : Float (default = 0.13): The maximum the mean of all channels’ drift can be before the tube is flagged as needing to be cleaned.

Segment Cutoff (default = 0.05): The minimum sum-of-measures’ CDF to keep.

Detect Worst Channels (Range) (default = 0): Try to detect the worst channels and use them to flag tubes / trim events. If this attribute is greater than 0, choose channels using the range of the mean of the bins’ fluorescence distribution. Often used in combination with Detect Worst Channels (SD).

Detect Worst Channels (SD) (default = 0): Try to detect the worst channels and use them to flag tubes / trim events. If this attribute is greater than 0, choose channels using the standard deviation of the mean of the bins’ fluorescence distribution. Often used in combination with Detect Worst Channels (Range).

Measures (default = ("5th percentile", "20th percentile", "50th percentile", "80th percentile", "95th percentile", "mean", "variance", "skewness") ).: Which measures should be considered when comparing segments?

Don't Clean (default = False): If True, never clean – just remove low-density bins.

Force Clean (default = False): If True, force cleaning even if the tube passes the original quality checks. Remember, the operation always removes low-density bins.