alt_fig5_network.Rmd

---
title: "Figure 5 - network analysis"
author: "Sara Gosline"
date: "2023-07-27"
output: html_document
---


```{r setup, include=FALSE}
knitr::opts_chunk$set(echo = TRUE)

library(ggplot2)
library(ggfortify)
library(cowplot)
#library(leapR)
library(dplyr)

source('spleenDataFormatting.R')
#library(spammer)
source('spatialProtUtils.R')
```

The assignment of pulp type from the basic data was not robust. Here we will use the cell type signatures from the sorted data to label the voxels.


## Cell type signatures

### using sorted cells
Here we get the differential expression from the sorted cells, and try to map them to spatial

```{r cell type signatures}


##get fit values for diffex
pumap<-scater::runPCA(spat.prot)

phumap<-spat.phos%>%
  scater::runPCA()
##proteomics data
spatDiff<-spatialDiffEx(pumap,column='Histology',vals=c('red','white'),'Protein')|>
   rowData(.)%>%
  as.data.frame()%>%
  dplyr::select(featureID='Protein',
                logFC='Histology.limma.logFC',
                adj.P.Val='Histology.limma.adj.P.Val',
                pval='Histology.limma.P.Value',
                AveExpr='Histology.limma.AveExpr')|>
  mutate(Signif=adj.P.Val<0.05,LogAdjP=log10(adj.P.Val)*-1)

##phosphoproteomics
phosDiff<-spatialDiffEx(phumap,column='Histology',vals=c('red','white'),'X')|>
   rowData(.)%>%
  as.data.frame()%>%
  dplyr::select(featureID='X',
                logFC='Histology.limma.logFC',
                adj.P.Val='Histology.limma.adj.P.Val',
                pval='Histology.limma.P.Value',
                AveExpr='Histology.limma.AveExpr')|>
  mutate(Signif=adj.P.Val<0.05,LogAdjP=log10(adj.P.Val)*-1)

diffExUpRegProts<-spatDiff|>
  subset(pval<0.05)|>
  subset(logFC>0)

print(diffExUpRegProts)

diffExUpRegPhos<-phosDiff|>
  subset(pval<0.1)|>
  subset(logFC>0)

```

now we have various choices of how we run the networks

```{r run networks}

all.phos<-phosDiff$logFC
names(all.phos)<-stringr::str_replace(phosDiff$featureID,'_','-')

sig.phos<-diffExUpRegPhos$logFC
names(sig.phos)<-stringr::str_replace(diffExUpRegPhos$featureID,'_','-')

all.prot<-spatDiff$logFC
names(all.prot)<-spatDiff$featureID

sig.prot<-diffExUpRegProts$logFC
names(sig.prot)<-diffExUpRegProts$featureID

protNet<-buildNetwork(sig.prot.vals=sig.prot,
                       sig.phos.vals=c(),
                       all.prot.vals=all.prot,
                       all.phos.vals=c(),
                       beta=10,nrand=1000,featName='whitePulpProt')

phosNet<-buildNetwork(sig.prot.vals=c(),
                       sig.phos.vals=sig.phos,
                       all.prot.vals=all.prot,
                       all.phos.vals=all.phos,
                       beta=5,nrand=1000,featName='whitePulpPhos')


combNet<-buildNetwork(sig.prot.vals=sig.prot,
                       sig.phos.vals=sig.phos,
                       all.prot.vals=all.prot,
                       all.phos.vals=all.phos,
                       beta=5,nrand=1000,featName='whitePulpAll')

nodes<-names(V(combNet$graph))
#V(combNet$graph)

##now we save the node information
##save node infor
```

The network approach was able to identify known white pulp proteins that were not in the global data (CD209, CD19, CD44, CD4)

```{r node stats}

node_stats<-readr::read_csv('whitePulpPhosnetwork.gml default node.csv')

p<-node_stats|>
  subset(BetweennessCentrality>0.05)|>
  ggplot(aes(x=reorder(name,BetweennessCentrality),y=BetweennessCentrality,fill=logFoldChange))+
  geom_bar(stat='identity')+
  theme(axis.text.x = element_text(angle = 45, hjust = 1))

 ggsave('centrality.pdf',p)
 
 p
```