Untangling statistical and biological models to understand network inference: the need for a genomics network ontology

Abstract

In this paper, we shed light on approaches that are currently used to infer networks from gene expression data with respect to their biological meaning. As we will show, the biological interpretation of these networks depends on the chosen theoretical perspective. For this reason, we distinguish a statistical perspective from a mathematical modeling perspective and elaborate their differences and implications. Our results indicate the imperative need for a genomic network ontology in order to avoid increasing confusion about the biological interpretation of inferred networks, which can be even enhanced by approaches that integrate multiple data sets, respectively, data types.

Post-inference Methods of Prior Knowledge Incorporation in Gene Regulatory Network Inference

“…It is reported that the GRN inference from the gene expression data infers only an influence network which contains transcription regulatory interactions along with other regulatory and non-regulatory interactions (Emmert-Streib et al, 2014;Matos Simoes et al, 2013;Hecker et al, 2009;Gardner and Faith, 2005). However, while quantifying the accuracy of the network inference, only the transcriptional regulatory interactions are considered as correct edges or true positives.…”

“…The dependencies between the mRNA transcripts can be due to interactions at transcription, posttranscription, protein, and metabolite (Emmert-Streib et al, 2014;Friedman, 2004;Gardner and Faith, 2005) which are not measured separately. Thus, the non-transcriptional interactions can be inferred due to the information limitation in the data.…”

“…Thus, the non-transcriptional interactions can be inferred due to the information limitation in the data. It is reported that the GRN inference methods based on the gene expression data can infer only an influence network (Emmert-Streib et al, 2014;Matos Simoes et al, 2013;Hecker et al, 2009;Gardner and Faith, 2005). The influence networks can contain any interaction that influences the mRNA expression levels such as the direct and indirect transcriptional regulatory interactions (Friedman, 2004); or other non-transcriptional regulatory and non-regulatory interactions such as the protein-protein interactions and metabolite interactions (Gardner and Faith, 2005).…”

“…Other factors such as noisy, limited, and incomplete data or errors in experiments can cause spurious interactions to be inferred (Emmert-Streib et al, 2014;Gardner and Faith, 2005). Thus, due to the limitations in technology or experimental errors, the inference may not capture the actual biological interaction or may predict spurious interactions.…”

“…Since, the prior knowledge is incorporated after the inference, any GRN inference method can be used. Further, it is reported that the methods of reverse engineering the GRN can predict only influence networks (Emmert-Streib et al, 2014;Matos Simoes et al, 2013;Hecker et al, 2009;Gardner and Faith, 2005) because the actual biological interactions cannot be differentiated from the indirect relations between genes. With the post-inference step of prior knowledge inclusion, it could be possible to differentiate the indirect relations from the actual interactions.…”