Skip to main content

Table 1 MA, DM and baseline methods included in the experimentations. For each method a synthetic description is provided describing its main characteristics

From: A comparative evaluation of data-merging and meta-analysis methods for reconstructing gene-gene interactions

Approach Method Description
Meta-Analysis Fisher Combines p-values in a statistic that follows a χ 2 distribution.
Stouffer Transforms p-values in Z-scores and merges them with a weighted average
Fixed-Effects Assumes all studies measure the same effect and combines estimates with a weighted average
Random-Effects Combines estimated effects by assuming that each study measures a biased version of the true effect
FR-Effects Estimates whether Fixed or Random-Effects assumptions hold and use one of the two methods accordingly
Rank-Product Combines statistics’ ranks by multiplication.
Data-Merging SVA Provides surrogate variables that approximate the effect of confounding factors and batch-effects present in the data
Combat Assumes additive and multiplicative batch-effects and estimates them by pooling information across genes
RMA Normalizes data across expression profiles using Quantile Normalization
RMA-Combat Applies RMA and Combat one after the other
Scaling Scales the value of each gene in each study to have zero mean and unitary standard deviation
No-Correction Merges samples from all studies in a single dataset without any correction
Baseline Single-Datasets Computes the performance that is expected by analyzing a single, randomly chosen dataset
Random-Guessing Produces randomly sampled correlation values