When to Use Each TraceBuildingStrategies

In order to multiplex image-based transcriptomics assays beyond the number of spectrally distinct fluorophores, assays use multiple rounds of imaging. Before every round, RNA transcripts or amplicons are relabeled so that if you trace a spot across rounds, the spot will have a pattern of signals. This pattern should match a codeword in the codebook.

There are three different TraceBuilders that can be used to trace spots in SpotFindingResults into an IntensityTable or DecodedIntensityTable. It is important to choose the correct TraceBuilder that matches the codebook design and data.

Which TraceBuilder To Use

The chosen TraceBuilder must also be compatible with how the SpotFindingResults was generated. FindSpotsAlgorithms can be run with or without a reference_image. If run with a reference_image then every PerImageSliceSpotResults in SpotFindingResults will have the same spots for every (round, channel) image volume. This is necessary for build_spot_traces_exact_match() but not recommended for build_traces_sequential() and build_traces_nearest_neighbors().




Reference Image


Build traces for every detected spot by setting intensity values to zero for all rounds and channels the spot was not found in (i.e. every trace will have only 1 non-zero value)



Build traces by combining intensity values of spots from every rounds and channel in the exact same location as spots in reference_image



Build traces by combining intensity values of spots from rounds and channels nearest to the spots in the anchor_round

Not recommended; will have same result as EXACT_MATCH

# Load and process ISS images to find spots with and without reference image
from starfish.image import ApplyTransform, LearnTransform, Filter
from starfish.types import Axes
from starfish import data, display, FieldOfView
from starfish.spots import FindSpots

experiment = data.ISS()
fov = experiment.fov()
imgs = fov.get_image(FieldOfView.PRIMARY_IMAGES) # primary images
dots = fov.get_image("dots") # reference round for image registration

# filter raw data
masking_radius = 15
filt = Filter.WhiteTophat(masking_radius, is_volume=False)
filt.run(imgs, in_place=True)
filt.run(dots, in_place=True)

# register primary images to reference round
learn_translation = LearnTransform.Translation(reference_stack=dots, axes=Axes.ROUND, upsampling=1000)
transforms_list = learn_translation.run(imgs.reduce({Axes.CH, Axes.ZPLANE}, func="max"))
warp = ApplyTransform.Warp()
warp.run(imgs, transforms_list=transforms_list, in_place=True)

# run blob detector on dots and on image stack
bd = FindSpots.BlobDetector(
dots_max = dots.reduce((Axes.ROUND, Axes.ZPLANE), func="max")
spots_from_ref = bd.run(image_stack=imgs, reference_image=dots_max)
spots_from_stack = bd.run(image_stack=imgs)

Typical pipelines will set the trace_building_strategy as an argument in the DecodeSpotsAlgorithm but here the underlying code is exposed to reveal what the different IntensityTables look like depending on which TraceBuilder is used.

from starfish.core.spots.DecodeSpots.trace_builders import build_spot_traces_exact_match, \
    build_traces_sequential, build_traces_nearest_neighbors

print('Build trace with EXACT_MATCH')

When building spot traces with EXACT_MATCH, every feature has a value in each round and channel because a reference_image was used in spot finding.

print('\nBuild trace with SEQUENTIAL')

When building spot traces with SEQUENTIAL, every feature has only one non-zero round and channel because build_traces_sequential() automatically assigns a zero value to all other rounds and channels.

print('\nBuild trace with NEAREST_NEIGHBORS')
print(build_traces_nearest_neighbors(spots_from_stack, search_radius=5))

When building spot traces with NEAREST_NEIGHBORS on spots found in ImageStack without a reference image, the nan values are due to no spot being found within the search_radius of the anchor_round.

print('\nBuild trace with NEAREST_NEIGHBORS')
print(build_traces_nearest_neighbors(spots_from_ref, search_radius=5))

The same build_traces_nearest_neighbors() applied to spots found in ImageStack with a reference image guarantees a spot to be found in every round of SpotFindingResults.

Gallery generated by Sphinx-Gallery