# **Network Biology approach to study miRNA-gene networks in human disorders**

###### tags: `science` `thesis` `Systems biology`
## **1. Abstract**
With the increasing complexity of interaction data being generated, diseases can better be understood in the context of network principles. This paper is an attempt to study the effect of aberrant expression and mutational consequences of the Rab GTPase family, on various Rab-associated disorders, and how the impact of defective genes can branch out along the links of the network and alter the activity of others via network biology. In modern-day biomedical research, multiple types of small RNAs like microRNAs have emerged to play a substantial role in biomarker discovery and potential therapeutic targets. Since the role of miRNAs in developmental and pathological processes is substantial, and given that miRNAs are found both in intracellular and extracellular level, miRNA-based studies can help us understand the pathobiology of Rab associated disorders, interconnecting these disease-causing Rabs to the miRNAs that regulate them at a cellular level.

### Keywords
###### *Rabs, network biology, miRNA, Rab associated diseases, enrichment analysis, network analysis*

## **2. Results and discussion**

We have made a comprehensive analysis by tabulating **87 diseases** that are also known to be associated with various members of the Rab family of GTPases. For this step, we have used DisGeNET, a database of genes and variants of human diseases, and tabulated the diseases along with the Rab reported to be associated with them, background Rabs and other affiliating genes involved. **Rab occupancy** for each disease sets was estimated which accounted for an average of **21.1409%** Rab occupancy in Rab-associated diseases.


###### **Figure 1:** Rab occupancy line-graph (Diseases numbered from 1-87)

---
Given that cellular components seldom act as single entities, rather act as interacting scaffolds, and that any biological interaction can be represented mathematically in the form of a graph $G= (V, E)$ where $V$ is the set of vertices or nodes and $E$ is the set of edges, it is indicative that the problem in question can be addressed by constructing miRNA-target gene co-expression networks for Rab-associated diseases.

**(a)** 
**(b)**
###### **Figure 2:** A force-atlas layout of miRNA- target gene co-expression networks of **(a)** Down Syndrome **(b)** Lupus Erythematosus, Systemic; generated using MiRNet. Nodes representing genes have been colored Red, miRNAs represented by small blue squares, and Rabs in enlarged pink circles. The network interactions (represented by edges) made by the Rabs in each of these diseases are highlighted in pink.
######

---
In all networks, most Rabs occupied **central positions**, showing that they play a major role in contributing to disease phenotypes. Since **Rabs** are members of a large family of **GTPases** comprising **72** known members to date, and given that Rabs are products of gene duplication, it is not surprising that Rabs involved in a disease tend to **cluster together**. It is therefore of certain interest to study these Rab sub-networks. Diseases sharing common genes hypothesize the presence of **common genetic origin**, which have been found in many cases in our study. Such genetic overlaps however have not been explicitly mentioned in this work since we are interested in network-based studies only.

---

Most biological networks are **scale-free**, and all networks in our study agree with the same. A scale-free topology essentially follows the **Power-law** and a **degree distribution** plot on a logarithmic scale shows that only a small number of nodes have a high degree and a large number of nodes have a low degree. However, real systems rarely follow a pure power law, and so is the case with miRNA-target gene networks. This deviation from the power law is marked by **low degree saturation** and a **high degree cut-off** in all of our networks. Whereas this phenomenon can be indicative of the fact a sub-population of biological networks are indeed **not purely scale-free**, not ignoring the fact the presence of latent phenomenon that is not clearly understood.

###### **Figure 3:** Scale-free property: miRNA-gene co-expression network in Polycystic ovary syndrome
######


###### **Figure 4:** Deviations from Power-law property: Low degree saturation and high degree cut-off in biological networks

---

**Computational deletion** of Rabs and their corresponding interactions established that most networks have remained **invariant** and **stable** even with the loss of links. However, invariability does not cover for a network’s vulnerability to targeted attack: if important links are broken off, the network loses connections and forms a set of isolated sub-networks. In our study, although the networks did not lose their overall connectivity, some very important links involved in crucial pathways might have been broken, thus imparting **disease symptoms**.


###### **Figure 5:** A degree distribution scatter-plot is shown (left) and a second scatter-plot showing the degree distribution after computational deletion of Rabs in the miRNA-gene co-expression network of Pancreatic Carcinoma

---

In most of the plots, we find a general trend: for nodes with higher degrees, the average clustering coefficient is low and for nodes with low degrees, the average clustering coefficient is high. A **trend analysis** was performed and the curve of the form $y=ab^x$ gave the best fit on the **clustering coefficient**, **closeness centrality**, and **Betweenness centrality** plots of the networks.

---