Skip to main content

Advertisement

Table 1 Measures of dependence between two variables

From: From protein-protein interactions to protein co-expression networks: a new perspective to evaluate large-scale proteomic data

Co-expression measures What measures? Input/Output Features
Pearson’s correlation (PC) Tendency to respond in opposite/same direction across different samples Input: gene expressions value
Output:
  • [0,1] both genes increase
  • [−1,0] one increase and other decrease
  • Sensitivity to outliers
  • Bad array of expression level can determine positive PC value
  • Measure linear relations
Spearman’s correlation (SC) Tendency to respond in opposite/same direction across different samples Input: ranking values from expression levels in samples
Output:
  • [0,1] Both genes increase
  • [−1,0] One increase and the other decrease
  • Robust to outliers
  • Detect non-linear associations
Mutual information Reduction of uncertainty of a gene given the knowledge about other gene Input: gene expression values
Output:
  • 0 there is no interdependence
  • >0 there is interdependence
  • Measure complex non-linear type relations (rarely present in biological data)
  • More samples are needed than PC, SC
  • Time-consuming computation
Kendall Correspondence/compatibility among two rankings Input: gene expression value
Output:
  • 1 perfect correspondence
  • -1 rankings exactly inverted
  • Similar to SC
  • Robust to outliers
  • Assumes fewer values than SC in the range [−1,1]