Algorithms for network-based identification of differential regulators from transcriptome data: a systematic evaluation

Abstract

Identification of differential regulators is critical to understand the dynamics of cellular systems and molecular mechanisms of diseases. Several computational algorithms have recently been developed for this purpose by using transcriptome and network data. However, it remains largely unclear which algorithm performs better under a specific condition. Such knowledge is important for both appropriate application and future enhancement of these algorithms. Here, we systematically evaluated seven main algorithms (TED, TDD, TFactS, RIF1, RIF2, dCSA_t2t, and dCSA_r2t), using both simulated and real datasets. In our simulation evaluation, we artificially inactivated either a single regulator or multiple regulators and examined how well each algorithm detected known gold standard regulators. We found that all these algorithms could effectively discern signals arising from regulatory network differences, indicating the validity of our simulation schema. Among the seven tested algorithms, TED and TFactS were placed first and second when both discrimination accuracy and robustness against data variation were considered. When applied to two independent lung cancer datasets, both TED and TFactS replicated a substantial fraction of their respective differential regulators. Since TED and TFactS rely on two distinct features of transcriptome data, namely differential co-expression and differential expression, both may be applied as mutual references during practical application.

Differential Regulatory Analysis Based on Coexpression Network in Cancer Research

“…For example, Ballouz et al suggest minimal experimental criteria to obtain useful functional connectivity and topology information of coexpression with microarrays greater than 20 samples and read depth greater than 10 M per sample [ 64 ]. Selection of differential regulatory genes relies on the method or algorithm chosen in DRA based on GCN [ 22 , 65 , 66 ]. For instance, differential analysis between two conditions, which is followed by regulatory analysis, can be performed after two gene coexpression networks are constructed.…”

ConGEMs: Condensed Gene Co-Expression Module Discovery Through Rule-Based Clustering and Its Application to Carcinogenesis

“…The gene co-expression network of identified genes will represent the blueprint of the inter-connections between them, along with the molecular regulators (i.e., transcription factors (TFs) and microRNAs) [ 28 ]. A gene co-expression module in network biology refers to a group of genes whose expression is highly related to the phenotype under examination and whose co-expression is highly related or significant [ 13 , 28 ]. From the literature, many measures regarding gene-module detection have already been proposed [ 29 ].…”

Enhanced identification of significant regulators of gene expression

“…There have also been a few examples of integration of RIF scores into other relevant software tools, like RMaNI [18], IN-sPeCT [19], DCGL [20], RegulatorTrail [21], and REGGAE [22]. There have been some testing and comparisons of RIF to other approaches, as for example in [23,24]. However, such comparisons can be challenging, at least partly because methods may have different requirements with respect to input data, e.g., gene lists vs. networks [24].…”

“…There have been some testing and comparisons of RIF to other approaches, as for example in [23,24]. However, such comparisons can be challenging, at least partly because methods may have different requirements with respect to input data, e.g., gene lists vs. networks [24].…”