Nonlinear Network Reconstruction from Gene Expression Data Using Marginal Dependencies Measured by DCOL

Abstract

Reconstruction of networks from high-throughput expression data is an important tool to identify new regulatory relations. Given that nonlinear and complex relations exist between biological units, methods that can utilize nonlinear dependencies may yield insights that are not provided by methods using linear associations alone. We have previously developed a distance to measure predictive nonlinear relations, the Distance based on Conditional Ordered List (DCOL), which is sensitive and computationally efficient on large matrices. In this study, we explore the utility of DCOL in the reconstruction of networks, by combining it with local false discovery rate (lfdr)–based inference. We demonstrate in simulations that the new method named nlnet is effective in recovering hidden nonlinear modules. We also demonstrate its utility using a single cell RNA seq dataset. The method is available as an R package at https://cran.r-project.org/web/packages/nlnet.

Dissecting cell identity via network inference and in silico gene perturbation

“…To test whether CellOracle can correctly identify cell-type-or cell-state-specific GRN configurations, we benchmarked our new method against diverse GRN inference algorithms: WGCNA, DCOL, GENIE3 and SCENIC. WGCNA is a correlation-based GRN inference algorithm, which is typically used to generate a non-directional network 53 ; DCOL is a ranking-based non-linear network modelling method 54 ; and GENIE3 uses an ensemble of tree-based regression models, and aims to detect directional network edges. GENIE3 emerged as one of the best-performing algorithms in a previous benchmarking study 55 .…”

Cell fate dynamics reconstruction identifies TPT1 and PTPRZ1 feedback loops as master regulators of differentiation in pediatric glioblastoma-immune cell networks

“…Network performance parameters epsilon (converge criteria threshold) and delta (binary detection threshold) were tuned to 1e-10 and 0.5, respectively. NLNET is a nonlinear hierarchical clustering, and variable selection algorithm using the distance based on conditional ordered list (DCOL) metric for identifying multi-nestedness (recursiveness) and community structures (modularity) within gene expression networks (Liu et al, 2016). The random permutations parameter for inferring the gene-specific null distribution was set to n=500, and the minimum and maximum cut-off false discovery rates were set to 0.05 and 0.2, respectively.…”

Gene Co-Expression Network Tools and Databases for Crop Improvement

“…To overcome this barrier, it has been suggested to combine linear and non-linear relationships, in which the GCN analysis results will be more comprehensive [73]. By using both approaches, multiple genes and higher-order regulatory patterns can be captured simultaneously and efficiently (i.e., regulatory interactions between transcription factors) [69].…”

“…Distance Correlation (DC) has also been used to measure non-linear relationships [75,76]. Although non-linear relationships are essential for complex interactions, they can be diverse, and the statistical power for detecting such relations is lower than linear-based correlation [73].…”