PMC

The workflow of REACTIN algorithm. Step 1: Measure of the binding affinity of a TF with all human genes using TIP. Step 2: Calculation of regulatory scores (RS) for all TFs. Step 3: Significance estimation and multiple-testing correction.

The activity difference of ER alpha between ER+ and ER- samples in all 10 datasets. Six ER alpha binding profiles are from T47d or Ecc1 cell lines, and treated with steroid hormone (Gen and Estradia) for 1h or Dmso2 as control. “*” indicates that a ER alpha binding profile has significantly high activity in ER+ than ER- with P<0.01.

The activity scores of six TF binding profiles for ER alpha in ER+ and ER-. The P-values in the top-right corner are calculated based on Wilcox rank sum test. The six TF binding profiles are from two cell lines (T47d and Ecc1) and under three different conditions (treated with steroid hormone Gen/Estradia for 1h, or with Dmso2 as control).

The relationship between survival time of patients with breast cancer and inferred E2F4 activity score. The Vijv dataset is used in the calculation. (a) Two breast cancer samples with an E2F4 regulatory score of -41.5 and 5.95, respectively. (b) Distribution of E2F4 activity scores in the 260 samples. (c) The survival curves of patients with breast cancer. “E2F4>0” shows patients with positive E2F4 activity scores; “E2F4<0” shows patients with negative E2F4 activity scores. (d) The survival curves of four categories patients: ER+ & E2F4>0, ER+ & E2F4<0, ER- & E2F4>0 and ER- & E2F4<0. E2F4 activity score is inferred based on the Sydh_Helas3_E2F4 binding profile.

REACTIN algorithm identifies significant activity difference of ER alpha (Haib_T47d_Eralphaa_Gen1h) in the Hess dataset. (a) Genes with higher t-scores (ER+ vs ER-) are more likely to be regulated by ER alpha. Genes are sorted in a decreasing order according to their t-scores (ER+ vs ER-). The –log10(P-value) is calculated by TIP, indicating the probability of a gene is bound by ER alpha in Haib_T47d_Eraphaa_Gen1h ChIP-seq data. The green lines indicates ER alpha target genes identified by peak-based method; (b) The correlation between the t-scores of genes and TF binding scores calculated by TIP; (c) The foreground and background functions for Haib_T47d_Eraphaa_Gen1h binding profile. The foreground and background functions are defined in Formula (xx) and (xx). Note the maximum deviation is obtained at the 18.9% percentile of all genes. (d) GSEA results for the ER alpha target gene sets defined by peak-based method (the green lines in (a)). Note that it cannot detect the activity difference of ER alpha between ER+ and ER- samples.