Comparison of statistical and optimisation-based methods for data-driven network reconstruction of biochemical systems

Abstract

Data-driven reconstruction of biological networks is a crucial step towards making sense of large volumes of biological data. Although several methods have been developed recently to reconstruct biological networks, there are few systematic and comprehensive studies that compare different methods in terms of their ability to handle incomplete datasets, high data dimensions and noisy data. The authors use experimentally measured and synthetic datasets to compare three popular methods - principal component regression (PCR), linear matrix inequalities (LMI) and least absolute shrinkage and selection operator (LASSO) - in terms of root-mean-squared error (RMSE), average fractional error in the value of the coefficients, accuracy, sensitivity, specificity and the geometric mean of sensitivity and specificity. This comparison enables the authors to establish criteria for selection of an appropriate approach for network reconstruction based on a priori properties of experimental data. For instance, although PCR is the fastest method, LASSO and LMI perform better in terms of accuracy, sensitivity and specificity. Both PCR and LASSO are better than LMI in terms of fractional error in the values of the computed coefficients. Trade-offs such as these suggest that more than one aspect of each method needs to be taken into account when designing strategies for network reconstruction.

Information theoretic approach to complex biological network reconstruction: application to cytokine release in RAW 264.7 macrophages

“…As Figure 8 shows, TNFα yields the best linear fit in terms of the coefficient of determination ( R 2 = 0.62), which is in good agreement with other models obtained by PCR [34,96] and PLS [44] methods. NF-κB p65 represents the highest statistical dependency while PKCμ2 has the lowest mutual information coefficient among the captured regulatory network components of GCS-F. JNK lg (from Toll data) shows the highest regulatory effect on IL-1α.…”

Time-Varying Causal Inference From Phosphoproteomic Measurements in Macrophage Cells

“…Type I error, Type II error, and accuracy of the network is computed [23] as follows using the False Positives (FP), False Negatives (FN), True Positives (TP) and True Negatives (TN) in the network identified: $$\text{italicType I Error}=\frac{\mathit{\text{FP}}}{\text{italicFP}+\text{italicTN}}$$ $$\text{italicType II Error}=\frac{\mathit{\text{FN}}}{\text{italicFN}+\text{italicTP}}$$ $$\text{italicAccuracy}=\frac{\text{italicTP}+\text{italicTN}}{\text{italicTP}+\text{italicTN}+\text{italicFP}+\text{italicFN}}.$$…”

Macrophages: Biology and Role in the Pathology of Diseases

“…Accordingly, several researchers have embarked on the development of methods for data-driven network reconstruction and their further integration and interpretation with legacy knowledge (Asadi et al 2012 ). Popular techniques for data-driven network reconstruction include regression and dimensionality-reduction based methods (e.g., statistical signifi cance tests combined with either principal components regressions (PCR), or partial least-squares (PLS) Pradervand et al 2006 ), partial-correlation-related analyses (Schafer and Strimmer 2005 ), Bayesian networks (Janes et al 2005 ;Sachs et al 2005 ), and hybrid methods such as Linear Matrix Inequalities (LMI) (Montefusco et al 2010 ) and Least Absolute Shrinkage and Selection Operator (LASSO) (Tibshirani 1996 ;Bonneau et al 2006 ).…”