Robust network inference using response logic

Abstract

Motivation:A major challenge in molecular and cellular biology is to map out the regulatory networks of cells. As regulatory interactions can typically not be directly observed experimentally, various computational methods have been proposed to disentangling direct and indirect effects. Most of these rely on assumptions that are rarely met or cannot be adapted to a given context. Results:We present a network inference method that is based on a simple response logic with minimal presumptions. It requires that we can experimentally observe whether or not some of the system's components respond to perturbations of some other components, and then identifies the directed networks that most accurately account for the observed propagation of the signal. To cope with the intractable number of possible networks, we developed a logic programming approach that can infer networks of hundreds of nodes, while being robust to noisy, heterogeneous or missing data. This allows to directly integrate prior network knowledge and additional constraints such as sparsity. We systematically benchmark our method on KEGG pathways, and show that it outperforms existing approaches in DREAM3 and DREAM4-challenges. Applied to a perturbation data set on PI3K and MAPK pathways in isogenic models of a colon cancer cell line, it generates plausible network hypotheses that explain distinct sensitivities towards EGFR inhibitors by different PI3K mutants. Availability and Implementation: A Python/Answer Set Programming implementation can be accessed at github.com/GrossTor/response-logic. Contact: nils.bluethgen@charite.de (Incomplete) transitive Closure Response Pattern Response logic Prior knowledge Logic Program Inference Rectification Union Conforming Networks Figure 1: The steps of the response logic approach. The response logic and all additional prior network knowledge are formulated as a logic program. It is first used to rectify the experimentally determined response pattern, and second, takes the resulting (potentially incomplete) transitive closure as input to infer either all individual conforming networks or the union thereof.work structure was resolved. Furthermore, we noticed that even though most network inference problems are embedded within very well studied contexts, the vast majority of reverse-engineering methods predicts networks de novo. Therefore, we additionally aimed for a method that could readily incorporate prior knowledge about presence or absence of certain links or about other known network properties. This resulted in what we call the response logic approach.In the following, we describe the response logic approach in more detail and then benchmark it by (i) assessing the performance using synthetic data derived from KEGG pathways [24], and (ii) comparing its performance to competing methods using the communitywide inference challenges (DREAM) [38]. Finally, we use the approach to study RAS/MAPK/PI3K signalling in a colon cancer cell line, and predict differences in the signalling network topology due to different PI3...

Perturbation graphs, invariant prediction and causal relations in psychology

“…More recently, Shojaie et al (2014) proposed an algorithm based on scoring rules from lasso regressions and order sorting algorithms; this algorithm is particularly suited for large-scale graphs where the only question is which nodes are connected but not the direction. Finally, a logic based algorithm was introduced (Gross et al, 2019) to cope with large networks and robust inference.…”

“…). This idea is used in what are called perturbation graphs (Klamt et al, 2010) or response graphs (Gross et al, 2019). These ideas described above are shown in Figure 1 for the two small graphs s → u → t and s ← u → t, with associated variables X s , X u and X t (see Appendix H for the R code and Appendix B for calculations of correlations).…”

scPerturb: Harmonized Single-Cell Perturbation Data

“…Large-scale single-cell perturbation-response screens enable exploration of complex cellular behavior not accessible from bulk observation. Directionality in regulatory network models cannot be inferred without interventional or time-series data about the system (Gross et al, 2019). Experiments with targeted perturbations can be modeled as affecting individual nodes of a regulatory network model, while the molecular readouts provide information on the state changes.…”

GAPORE: Boolean network inference using a genetic algorithm with novel polynomial representation and encoding scheme

“…A gene regulatory network is composed of interactions among the genes and proteins in living cells and reflects the relationship of these genes or proteins [26,33]. Gene regulatory networks play an increasingly crucial role in uncovering and analyzing the underlying regulatory mechanism of biological organisms from a systematic view [14,43,33,36]. There are many approaches to reconstruct gene regulatory networks by using time-series data from experimental observation, such as fuzzy cognitive maps [37,38], Bayesian networks [24], Boolean networks [4,39].…”