Drugst.One offers the possibility to analyze the displayed genes and proteins (nodes) in regard to network models and drug repurposing. Given nodes can be connected by shortest paths, modules can be extended by new highly interconnected nodes. Further, drugs can be found that target the nodes in the network. Thus, Drugst.One also functions as a drug repurposing tool.
To this end, Drugst.One offers different algorithms which can be chosen by the user to fit the application case. It is up to the developer of the website, which algorithms are available for selection or if the analysis function is enabled at all.
Algorithm | Gene Search | Drug Search |
---|---|---|
Multi-Level Steiner Trees | ✓ | X |
KeyPathwayMiner | ✓ | X |
Harmonic Centrality | ✓ | ✓ |
Degree Centrality | ✓ | ✓ |
Betweenness Centrality | ✓ | X |
TrustRank | ✓ | ✓ |
Network Proximity | X | ✓ |
The Multi-level Steiner tree algorithm is used to approximate the minimum spanning tree spanned by the seed nodes in a reasonable time. The implementation is adopted from Ahmed et al. It can be used to create a minimum spanning subnetwork between user-selected seed nodes, which happen to be central interaction partners between the seed nodes, and thus represent favorable drug-targets.
KeyPathwayMiner (KPM) is an online tool developed by Alcaraz et al. for pathway enrichment analysis. Users have the option to utilize KPM for their drug-target search by selecting seed nodes in the GGI network and letting KPM find an interaction network of genes that are commonly dysregulated spanned by the seed genes. The resulting proteins are presumably functionally related to the seed nodes, and therefore are suitable drug-target candidates. Only one parameter K has to be set by the user which defines the amount of permitted intermediate nodes that are neither part of the seed nodes nor the common pathway.
Harmonic centrality measurement can be described as the average shortest distance from each node to all other nodes in a network. This measurement is the equivalent of harmonic centrality for disconnected graphs. Formally speaking, it can be annotated as
where x is a given node and 1/dist(x, y) = 0 if dist(x, y) = ∞ (reference). The closer a node is to other nodes, the higher the score. It has already been proven successful in a number of biological network problems for instance with metabolic or PPI networks .
Degree centrality measurement is obtained by ranking the nodes in a network based on their degree, which is defined as the number of neighbors a node has divided by the number of nodes. It be described as
where x is a given node and deg(x) is its degree. While it is a commonly used network analysis technique, it most importantly has been shown useful in the identification of essential proteins in PPI networks. Thus, it is a simple approach to classifying the network-related importance of a particular protein. In CADDIE it can be used to discover valuable drug-targets or drugs, based on the seed selection given by the user.
Betweenness is obtained by finding the shortest paths for each pair of nodes in the network and assessing the number of shortest paths that pass through a particular node, such that a measure of the centrality of a node in a network global context is received. Betweenness Centrality has been established as a common measurement in network biological application and is especially practical in finding communities in large networks. In Drugst.One, betweenness is based on the shortest paths between the seed nodes only and can be used to find drug targets with maximized connectivity to all seeds.
TrustRank is based on the same concepts as the Google PageRank algorithm and Harmonic Centrality. A crawler searches the network based on user-selected seeds and ranks visited nodes, damping the score based on the distance traveled. The damping factor can be set by the user in a range from 0-1, with a higher damping factor causing the crawler to explore nodes in close proximity or in larger portions of the GGI network. In Drugst.One, TrustRank is able to find putative drug targets as well as drug candidates.
Network Proximity, as introduced by Guney et al., is the average length of shortest paths from drug target nodes to all of the user-selected seed nodes. The algorithm then computes a statistical significance score comparing against random expectation. This algorithm was adopted in Drugst.One so that best-scored drugs are returned to the user as candidate drugs.
All of the algorithms can be adjusted to your needs with parameters. We also provide custom options such as filtering out unapproved or nutraceutical drugs. In general, the following options exist: