Abstract
Reverse engineering of gene regulatory networks (GRNs) is a central task in systems biology. Most of the existing methods for GRN inference rely on gene co-expression analysis or TF-target binding information, where the determination of co-expression is often unreliable merely based on gene expression levels, and the TF-target binding data from high-throughput experiments may be noisy, leading to a high ratio of false links and missed links, especially for large-scale networks. In recent years, the microscopy images recording spatial gene expression have become a new resource in GRN reconstruction, as the spatial and temporal expression patterns contain much abundant gene interaction information. Till now, the spatial expression resources have been largely underexploited, and only a few traditional image processing methods have been employed in the image-based GRN reconstruction. Moreover, co-expression analysis using conventional measurements based on image similarity may be inaccurate, because it is the local-pattern consistency rather than global-image-similarity that determines gene-gene interactions. Here we present GripDL (Gene regulatory interaction prediction via Deep Learning), which incorporates high-confidence TF-gene regulation knowledge from previous studies, and constructs GRNs for Drosophila eye development based on Drosophila embryonic gene expression images. Benefitting from the powerful representation ability of deep neural networks and the supervision information of known interactions, the new method outperforms traditional methods with a large margin and reveals new intriguing knowledge about Drosophila eye development.
Highlights
Over the past decades, the advances of high-throughput technologies have led to a rapid accumulation of genomic, transcriptomic, proteomic and metabolomics data, and enabled the studies of gene regulation and gene-gene interactions at genome scale [1, 2]
We demonstrate its performance by inferring large-scale gene regulatory networks (GRNs) for Drosophila eye development based on spatial expression patterns of Drosophila embryos
The abundance of spatial expression data has enabled the inference of gene regulatory networks based on spatial distribution of gene expression, and revealed a lot of new regulatory associations that are undetected by traditional experiments
Summary
The advances of high-throughput technologies have led to a rapid accumulation of genomic, transcriptomic, proteomic and metabolomics data, and enabled the studies of gene regulation and gene-gene interactions at genome scale [1, 2]. The reverse engineering algorithms for GRNs aim to identify edges between nodes so as to infer the network structure, where the edges (interactions) have two major types, i) physical/direct interactions, i.e., interactions between transcription factors (TFs) and their target genes, usually revealed by ChIPChip or ChIP-Seq experiments; ii) influential/indirect interactions (i.e. gene interaction network, GIN), inferred by similar expression levels from DNA microarray or next-generation sequencing profiles [4]. The identification of both types has attracted a lot of research interests [5], though there may be not a clear distinction between GRN and GIN. Puniyani et al provided an example (Fig 1 in [11]), in which two genes have completely different spatial expression patterns over time, but their averaged values are nearly identical, suggesting that the averaging operation would lead to unreliable results
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
