CellTracksColab - breast cancer cell dataset

Dataset used in the manuscript "CellTracksColab—A platform for compiling, analyzing, and exploring tracking data"

This Zenodo archive contains:

• The raw video (Raw zip files)
• The tracking files as XML and CSV files (Tracks.zip)
• The masks used to identify the edges of the monolayer (Monolayer_edges.zip)
• The CellTracksColab dataframes storing the dataset (CellTracksColab_results.zip)
• The CellTracksColab outputs used to make the figures in the paper (CellTracksColab_results.zip)

In brief:

In this experiment, approximately 50,000 shCTRL or shMYO10 lifeact-RFP DCIS.COM cells were seeded into one well of an ibidi culture-insert 2 well pre-placed in a µ-Slide 8 well. The cells were cultured for 24 hours, after which the culture insert was removed to create a wound-healing assay setup. When appropriate, a fibrillar collagen gel (PureCol EZ Gel) was applied over the cells and allowed to polymerize for 30 minutes at 37°C. Standard culture media was added to all wells, and the cells were left to migrate/invade for two days. Before live cell imaging, the cells were treated with 0.5 µM SiR-DNA (SiR-Hoechst, Tetu-bio) for two hours. Imaging was performed over 14 hours using a Marianas spinning-disk confocal microscope system. This system included a Yokogawa CSU-W1 scanning unit mounted on an inverted Zeiss Axio Observer Z1 microscope (Intelligent Imaging Innovations, Inc.). Imaging was conducted using a 20x (NA 0.8) air Plan Apochromat objective (Zeiss), and images were captured at 10-minute intervals.

Cell tracking was conducted using Fiji and TrackMate. The Stardist detector was employed to detect nuclei using the Stardist versatile model. Tracks were created using the Kalman tracker (a maximum frame gap of 1, a Kalman search radius of 20 µm, and a linking maximum distance of 15 µm). Post-tracking, tracks were filtered so that each track had to contain more than six spots, ensuring a significant amount of data per track, and the total distance traveled by cells had to be greater than 89 µm.

In CellTracksColab, we conducted a dimensionality reduction analysis employing Uniform Manifold Approximation and Projection (UMAP). The UMAP settings were as follows: number of neighbors (n_neighbors) set to 10, minimum distance (min_dist) to 0.5, and number of dimensions (n_dimension) to 2. This analysis utilized an array of track metrics, including:

NUMBER_SPOTS, NUMBER_GAPS, NUMBER_SPLITS, NUMBER_MERGES, NUMBER_COMPLEX, LONGEST_GAP, TRACK_DURATION, TRACK_DISPLACEMENT, TRACK_MEAN_SPEED, TRACK_MAX_SPEED, TRACK_MIN_SPEED, TRACK_MEDIAN_SPEED, TRACK_STD_SPEED, TRACK_MEAN_QUALITY, TOTAL_DISTANCE_TRAVELED, MAX_DISTANCE_TRAVELED, CONFINEMENT_RATIO, MEAN_STRAIGHT_LINE_SPEED, LINEARITY_OF_FORWARD_PROGRESSION, MEAN_DIRECTIONAL_CHANGE_RATE, Directionality, Tortuosity, Total_Turning_Angle, Spatial_Coverage, MEAN_RADIUS, MEAN_CIRCULARITY, MEAN_SOLIDITY, MEAN_SHAPE_INDEX, MEDIAN_RADIUS, MEDIAN_CIRCULARITY, MEDIAN_SOLIDITY, MEDIAN_SHAPE_INDEX, STD_RADIUS, STD_CIRCULARITY, STD_SOLIDITY, STD_SHAPE_INDEX, MIN_RADIUS, MIN_CIRCULARITY, MIN_SOLIDITY, MIN_SHAPE_INDEX, MAX_RADIUS, MAX_CIRCULARITY, MAX_SOLIDITY, MAX_SHAPE_INDEX, MaxDistance_edge, MinDistance_edge, StartDistance_edge, EndDistance_edge, MedianDistance_edge, StdDevDistance_edge, DirectionMovement_edge, AvgRateChange_edge, PercentageChange_edge, TrendSlope_edge.

Subsequently, clustering analysis was performed using Hierarchical Density-Based Spatial Clustering of Applications with Noise (HDBSCAN). The parameters included clustering_data_source set to UMAP, min_samples at 20, min_cluster_size at 200, and the metric employed was Canberra.