Abstract
Recent developments in large-scale monitoring of gene expression such as DNA microarrays have made the reconstruction of gene regulatory networks (GRNs) feasible. Before one can infer the structures of these networks, it is important to identify, for each gene in the network, which genes can affect its expression and how they affect it. Most of the existing approaches are useful exploratory tools in the sense that they allow the user to generate biological hypotheses about transcriptional regulations of genes that can then be tested in the laboratory. However, the patterns discovered by these approaches are not adequate for making accurate prediction on gene expression patterns in new or held-out experiments. Therefore, it is difficult to compare performance of different approaches or decide which approach is likely to generate plausible hypothesis. For this reason, we need an approach that not only can provide interpretable insight into the structures of GRNs but also can provide accurate prediction. In this paper, we present a novel fuzzy logic-based approach for this problem. The desired characteristics of the proposed algorithm are as follows: (i) it is able to directly mine the high-dimensional expression data without the need for additional feature selection procedures, (ii) it is able to distinguish between relevant and irrelevant expression data in predicting the expression patterns of predicted genes, (iii) based on the proposed objective interestingness measure, no user-specified thresholds are needed in advance, (iv) it can make explicit hidden patterns discovered for possible biological interpretation, (v) the discovered patterns can be used to predict gene expression patterns in other unseen tissue samples, and (vi) with fuzzy logic, it is robust to noise in the expression data as it hides the boundaries of the adjacent intervals of the quantitative attributes. Experimental results on real expression data show that it can be very effective and the discovered patterns reveal biologically meaningful regulatory relationships of genes that could help the user reconstructing the underlying structures of GRNs.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have