Skip to contents

Initialize tile scores for aggregation

Source: R/utils_aggregation.R
init_scores.Rd

Higher scores favor merging.

Usage

init_scores(aggs, agg_mode, ...)

Arguments

aggs

A tile data structure.

agg_mode

Method to use to calculate aggregation scores (1, 2, or 3).

...

Additional parameters for different modes:

  1. No additional parameters.

  2. Requires alpha and max_npts. For alpha, 0.2 = conservative merging, 2 = liberal merging.

  3. Requires alpha and min_npts. For alpha, 0.2 = conservative merging, 2 = liberal merging.

Value

aggs with scores for aggregation stored in attributes:

edges

Additional attributes are calculated: * dscore: Overall score for merging two tiles. Product of w, score_size, and dC. * w: Gene expression similarity score. * score_size: Penalizes tiles with many points. * perimeter_merge: Perimeter of merged tile.

d_mu,d_sig

Parameters used to calculate w.

pcs_merged

Average PCs for merged tile.

Details

Methods for different modes of aggregation:

  1. dscore has already been manually calculated and stored in aggs$edges$dscore. Will set the score, score_size, and compactness attributes of aggs$meta_data to 0.

  2. dscore is the product of three factors:

    • w: A 2-cluster GMM is used to determine the mean mu and standard deviation sig of the distance between tiles that have similar gene expression (Euclidean distance d in PC space). Then we define d_mu = mu + sig and d_sig = alpha * sig, and calculate w as w = 0.5 - 1 / (1 + exp(-(d - d_mu) / d_sig)). Ranges from -0.5 to 0.5. If adjacent tiles are very dissimilar (d >> d_mu), then d is large, and w is close to -0.5. If adjacent tiles are very similar (d < d_mu), then d is small, and w is positive.

    • score_size: (1 - npts_from/max_npts) * (1 - npts_to/max_npts). Ranges from 0 to 1. If merging the two tiles would have a total number of points ≥max_npts, then score_size is -Inf, which prevents merging.

    • dC: .5 * (C_merge - C_from - C_to + 1). Ranges from 0 to 1.

  3. dscore is the product of same three factors as mode 2, but score_size has already been precomputed. Additionally, dscore is set to -1 if both adjacent tiles have at least min_npts cells.