calculate_autocorrelation: Calculate the autocorrelation for each protein between a pair...

In fosterlab/PrInCE: Predicting Interactomes from Co-Elution

Description

For a given protein, the correlation coefficient to all other proteins in the first condition is calculated, yielding a vector of correlation coefficients. The same procedure is repeated for the second condition, and the two vectors of correlation coefficients are themselves correlated, yielding a metric whereby higher values reflect proteins with unchanging interaction profiles between conditions, while lower values reflect proteins with substantially changing interaction profiles.

Usage

calculate_autocorrelation(
  profile1,
  profile2,
  cor_method = c("pearson", "spearman", "kendall"),
  min_replicates = 1,
  min_fractions = 1,
  min_pairs = 0
)

Arguments

profile1	a numeric matrix or data frame with proteins in rows and fractions in columns, or a MSnSet object, representing the first co-elution condition
profile2	a numeric matrix or data frame with proteins in rows and fractions in columns, or a MSnSet object, representing the second co-elution condition
cor_method	the correlation method to use; one of "pearson", "spearman", or "kendall").
min_fractions	filter proteins not quantified in at least this many fractions
min_pairs	remove correlations between protein pairs not co-occuring in at least this many fractions from the autocorrelation calculation

Details

Note that all of zero, NA, NaN, and infinite values are all treated equivalently as missing values when applying the min_fractions and min_pairs filters, but different handling of missing values will produce different autocorrelation scores.

Value

a named vector of autocorrelation scores for all proteins found in both matrices.

fosterlab/PrInCE documentation built on Dec. 13, 2020, 5:50 a.m.