Monthly Archives: October 2016

A global genetic interaction network maps a wiring diagram of cellular function

In our last group meeting, I talked about a recent paper which presents a vast amount of genetic interaction data as well as some spatial analysis of the created data. Constanzo et al. used temperature-sensitive mutant alleles to measure the interaction of ~6000 genes in the yeast Saccharomyces cerevisiae [1]. A typical way to analyse such data would be the use of community detection to find groups of genes with similar interaction pattern, see for example [2] for a review. Instead, the authors of this paper created a two-dimensional embedding of the network with a spring-layout, which places nodes close to each other if they show similar interaction pattern.

The network layout is then compared with Gene Ontology by applying a spatial analysis of functional enrichment (SAFE) [3]. Clusters enriched are associated for example with cell polarity, protein degradation, and ribosomal RNA. By filtering the network for different similarities they find a hierarchical organisation of genetic function with small dense modules of pathways or complexes at the bottom and sparse clusters representing different cell compartments at the top.

In this extensive paper, they then go further into detail to quantify gene pleiotropy, predict gene function, and how the interaction structure differs between essential and non-essential genes. They also provide that data online under http://thecellmap.org/costanzo2016/ .

(Left) A network of genetic interaction was embedded into a two-dimensional space using a spring-layout (Right) This embedding was compared with Gene Ontology terms to find regions of spatial enrichment.
Image from [1]

References:

[1] Costanzo, Michael, et al. “A global genetic interaction network maps a wiring diagram of cellular function.” Science 353.6306 (2016): aaf1420.

[2] Fortunato, Santo. “Community detection in graphs.” Physics Reports 486.3 (2010): 75-174.

[3] Baryshnikova, Anastasia. “Systematic Functional Annotation and Visualization of Biological Networks.” Cell Systems (2016).

Viewing 3D molecules interactively in Jupyter iPython notebooks

Greg Landrum, curator of the invaluable open source cheminformatics API, RDKit, recently blogged about viewing molecules in a 3D window within a Jupyter-hosted iPython notebook (as long as your browser supports WebGL, that is).

The trick is to use py3Dmol. It’s easy to install:

pip install py3Dmol

This is built on the object-oriented, webGL based JavaScript library for online molecular visualization 3Dmol.js (Rego & Koes, 2015); here's a nice summary of the capabilities of 3Dmol.js. It's features include:

support for pdb, sdf, mol2, xyz, and cube formats
parallelized molecular surface computation
sphere, stick, line, cross, cartoon, and surface styles
atom property based selection and styling
labels
clickable interactivity with molecular data
geometric shapes including spheres and arrows

I tried a simple example and it worked beautifully:

import py3Dmol
view = py3Dmol.view(query='pdb:1hvr')
view.setStyle({'cartoon':{'color':'spectrum'}})
view

The 3Dmol.js website summarizes how to view molecules, along with how to choose representations, how to embed it, and even how to develop with it.

References

Nicholas Rego & David Koes (2015). “3Dmol.js: molecular visualization with WebGL”.
Bioinformatics, 31 (8): 1322-1324. doi:10.1093/bioinformatics/btu829

SMEs in Research

Scientific research relies on collaboration, between academics across the world, between big and small groups of people, and between companies and universities. Many people’s first thoughts of companies involved in academia will be large multinational corporations, such as pharmaceutical or aerospace companies. However mutually beneficial relationships exist between smaller companies and academia. Scientific spin out companies, such as Oxford Nanopore Technologies, Theta Technologies or Reinnervate, are one way research is expanded from an idea held by researchers at a university and expanded into a commercial product or service. However, there are many other ways and reasons for scientists collaborating with small companies.

Small and medium sized enterprises (SMEs) are independent businesses with up to 249 people, they represent over 99% of all UK and EU businesses, and 45-50% of turnover and employment. But why would researchers be interested in involvement with SMEs? Access to unique intellectual property and innovation in the SMEs, as well as access to funding targeted at fostering research led projects in SMEs. Innovate UK, support growth by enabling and funding innovative opportunities. SMEs can access this support through Knowledge Transfer Partnerships (KTP), a three-way partnership between a graduate, academic institution and a business. These KTPs can last between 12 months and 3 years, and provide financial support for academics to monitor the graduate’s work. KTP projects lead to an average of 2 papers per project, and are attractive to SMEs as they only fund 33% of the project cost.

Impact of research outside academia is becoming increasingly important to identify as impact case studies form a part of the Research Excellence Framework (REF), making many sources of funding contingent on providing a strong case for the wider impact of the research. Work with innovation focused SMEs can provide a focus for research impact in engagement and economic terms. For example, a REF case study highlighting the impact of spintronics research in the development of non-contact sensors, through a spin out company, Salunda.

Plotting and storing a 3D network in R

A simple toy example of a three layered network:

Note 1: In order to view the 3D plots, mac users will need Xquartz installed (https://www.xquartz.org/).

 
require(igraph)
require(rgl)
#Another package that might be needed is "rglwidget". The function writeWebGL will show an error stating if rglwidget is required.
######################################//// 
######The basics######################////
######################################////
#1) Create a "food" network (three layers) 
set.seed(432)
g1<-watts.strogatz.game(dim = 1,size = 5,nei = 2,p = .5,loops = FALSE,multiple = FALSE)
g2<-watts.strogatz.game(dim = 1,size = 10,nei = 2,p = .2,loops = FALSE,multiple = FALSE)
g3<-watts.strogatz.game(dim = 1,size = 30,nei = 1,p = .5,loops = FALSE,multiple = FALSE)
g123=g1+g2+g3 


#Create more edges btw layers 
g123=rewire(g123,each_edge(prob=.4,loops = FALSE,multiple = FALSE)) 
ne=15;add_edges(g123,edges = cbind(sample(1:vcount(g1),size = ne,replace = TRUE), sample((vcount(g1)+1):vcount(g123),size = ne,replace = TRUE)))#top layer 
ne=30;add_edges(g123,edges = cbind(sample((vcount(g1)+1):(vcount(g1)+vcount(g2)),size = ne,replace = TRUE), sample((vcount(g1)+vcount(g2)+1):vcount(g123),size = ne,replace = TRUE)))#second layer 

#A quick plot of the graph
plot(g123,vertex.size=1,vertex.label.cex=0.02)


#Create 3d coordinates of the network layout
circpos=function(n,r=1){#Coordinates on a circle
 rad=seq(0,2*pi,length.out=n+1)[-1];x=cos(rad)*r;y=sin(rad)*r
 return(cbind(x,y))
}
#
lay=rbind(cbind(circpos(vcount(g1),r=1), runif(n = vcount(g1),-1,1)),
 cbind(circpos(vcount(g2),r=2), runif(n = vcount(g2),6,7)),
 cbind(circpos(vcount(g3),r=4), runif(n = vcount(g3),13,17))
)



#2d plot using the previous layout
plot(g123,vertex.size=5,vertex.label=NA,layout=lay[,c(1,3)])
plot(g123,vertex.size=1,vertex.label=NA,layout=lay[,c(1,2)])

 
#3D graph plot
#Add some colour to nodes and edges
nodecols=c(rep("red",vcount(g1)),
 rep("blue",vcount(g2)),
 rep("yellow",vcount(g3)))

edgecols=function(elist,cols,grouplist){
 whatcol=rep(length(cols)+1,nrow(elist))
 finalcol=whatcol
 for(i in 1:nrow(elist)){
 for(k in length(cols):1){ 
 if( k * (length( intersect(elist[i,], grouplist[[k]]) ) > 0)){
 whatcol[i]=min(whatcol[i], k )
 }
 }
 finalcol[i]=cols[whatcol[i]]
 }
 return(finalcol)
}

#Open 3d viewer
rgl.open()
rglplot(g123, layout=lay,vertex.size=5,vertex.label=NA,vertex.color=nodecols,
 edge.color=edgecols(elist=get.edgelist(g123,names = FALSE),cols=c("orange","green","pink"),grouplist=list(1:vcount(g1), (vcount(g1)+1):(vcount(g1)+vcount(g2)), (vcount(g1)+vcount(g2)+1):vcount(g123)) )
)

###Storing the plot in an html file###

dirfolder="..." #your dir
#rgl.open()#instead of rgl.open use open3d, in order to save the plot. 
open3d()
rglplot(g123, layout=lay,vertex.size=5,vertex.label=NA,vertex.color=nodecols,
 edge.color=edgecols(elist=get.edgelist(g123,names = FALSE),cols=c("orange","green","pink"),grouplist=list(1:vcount(g1), (vcount(g1)+1):(vcount(g1)+vcount(g2)), (vcount(g1)+vcount(g2)+1):vcount(g123)) )
)
#Fix the view
rgl.viewpoint(theta=90, phi=0)

#Save a static 2d image:
rgl.snapshot(paste(dirfolder,"a_png_pic_0.png",sep=""), fmt="png", top=TRUE)

#Save the plot in a .htlm file:
rglfolder=writeWebGL(dir = paste(dirfolder,"first_net3d",sep=""), width=700)

#The previous function should create a file called index.htlm inside the folder "first_net3d". By opening this file in a browser (with javascript enabled) the 3d plot will be displayed again.
#Also the following command will open the plot in the browser:
browseURL(rglfolder)

Note 2: In order to view the .htlm file javascript should be enabled in the browser. (Here is an example on how to do this for safari ).

Although not covered in the previous script, further options are available such as edge/vertex size and the ability to control independently each of the nodes and edges in the graph. Here is an example that makes more use of these options:

3d network representing a T cell receptor. Edges are coloured according to a relevant path found between the bottom green node and the upper red node cluster.

T cell receptor (in blue), binding to a peptide (in red).

Is contacts-based protein-protein affinity prediction the way forward?

The binding affinity of protein interactions is useful information for a range of protein engineering and protein-protein interaction (PPI) network challenges. Obvious applications include the development of therapeutic antibodies to given drug targets or the engineering of novel interfaces for synthetic protein complexes. An accurate model would furthermore allow us to predict a large proportion of affinities in existing PPI networks, and enable the identification of new PPIs, which is critical for our ability to model protein network dynamics effectively.

“The design of an ideal scoring function for protein−protein docking that would also predict the binding affinity of a complex is one of the challenges in structural proteomics.” Adapted from Kastritis, Panagiotis L., and Alexandre MJJ Bonvin. Journal of proteome research 9.5 (2010): 2216-2225.

In last week’s paper a new binding-affinity prediction method based on interfacial contact information was described. Contacts have long been used to in docking methods but surprisingly this was the first time that binding affinity was predicted with them. Largely, this was due to the lack of a suitable benchmark data set that contained structural as well as affinity data . In 2011, however, Kastritis et al. presented a curated database of 144 non-redundant protein–protein complexes with experimentally determined Kd (ΔG) as well as x-ray structures.
Using this data set they trained and validated their method, compared it against others and concluded that interfacial contacts `can be considered the best structural property to describe binding strength`. This claim may be true but as we discussed in the meeting there is still some work to do before we take this model an run with it. A number of flags were raised:

Classification of experimental methods into reliable and non-reliable is based on what gives the best results with their method. Given that different types of protein complexes are often measured with different methods, some protein classes for which contact-based predictions are less effective may be excluded.
Number of parameters for model 6 is problematic without exact AIC information. As Lyuba righlty pointed out, the intercept in model 6 `explodes`. It is no surprise that the correlation improves with more parameters. Despite their AIC analysis, overfitting is still a worry due to the lack of details presented in the paper.

Comparison against other methods is biased in their favour; their method was trained on the same data set, the others were not. In order to ensure a fair comparison all methods should be trained on the same data set. Of course this is hard to do in practice, but the fact remains that a comparison of methods that has been trained on different data sets will be flawed.

Paper: Vangone, A., Bonvin, A. M. J. J., Alberts, B., Aloy, P., Russell, R., Andrusier, N., … Zhou, Y. (2015). Contacts-based prediction of binding affinity in protein-protein complexes. eLife, 4, e07454. http://doi.org/10.7554/eLife.07454

Oxford Protein Informatics Group

or "OPIG" to friends