figX_plexDistanceCorrelation.Rmd

---
title: "Plex Analysis"
author: "Sara Gosline"
date: "`r Sys.Date()`"
output: html_document
---

```{r setup, include=FALSE}
library(dplyr)
library(purrr)
library(devtools)
#library(MSnSet.utils)
if(!require(MSnSet.utils))
  devtools::install_github('PNNL-Comp-Mass-Spec/MSnSet.utils')

library(ggplot2)
library(ggfortify)
library(cowplot)
library(leapR)
source('loadHumanPancData.R')
```



## Plex analysis

Each plex is labeled with a heavy isotope that is not 100% pure, as such an isotope with a similar molecular weight can be detected in a different plex. As such, we use the difference in molecular weight to see if this correlates with similarity between plexes.


```{r labels}

print(tidyr::pivot_wider(labels,names_from='OtherPlex',values_from='mwDist'))
```

## Correlation between samples

We can now do a pairwise correlation between samples to see if we observe any bias in correlation between samples that share plexes. We verify the labels of the plexes and then map them, along with the distnaces, to the correlation values

```{r map plex labels to correlatioon values}

plexLabels <- isletMeta%>%
  tibble::rownames_to_column('Orig')%>%
  dplyr::select(Orig,Plex,IsletStatus,IsletOrNot,Image='Islet Number')%>%
  distinct()

print(plexLabels)

otherLabels <- plexLabels
names(otherLabels)<-paste0('Other',names(plexLabels))

norm.cor <- cor(crosstab,use='pairwise.complete.obs',method='spearman')%>%
  as.data.frame(row.names=colnames(crosstab))%>%
  tibble::rownames_to_column('Orig')%>%
  tidyr::pivot_longer(2:(ncol(crosstab)+1),names_to='OtherOrig',values_to='corVal')%>%
  left_join(plexLabels)%>%
  left_join(otherLabels)%>%
  left_join(labels)%>%
  mutate(labelStatus=ifelse(mwDist==0,'Same',ifelse(mwDist==1,'OneAWay','Other')))%>%
  mutate(ImageStatus=Image==OtherImage)%>%
  mutate(Image=as.factor(Image))

library(ggplot2)


ggplot(norm.cor,aes(x=mwDist,y=corVal,col=mwDist))+
  geom_jitter()+geom_boxplot()+ggtitle("Corrected pixel correlations")+facet_grid(Image~OtherImage)



ggplot(norm.cor,aes(x=mwDist,y=corVal,col=Image))+
  geom_jitter()+geom_boxplot()+ggtitle("Corrected pixel correlations")+facet_grid(Plex~OtherPlex)

ggplot(subset(norm.cor,ImageStatus),aes(x=mwDist,y=corVal,col=Image,))+
  geom_jitter()+geom_boxplot()+ggtitle("Corrected pixel correlations")


orig.cor <- cor(crosstab2,use='pairwise.complete.obs',method='spearman')%>%
  as.data.frame(row.names=colnames(crosstab))%>%
  tibble::rownames_to_column('Orig')%>%
  tidyr::pivot_longer(2:(ncol(crosstab)+1),names_to='OtherOrig',values_to='corVal')%>%
  left_join(plexLabels)%>%
  left_join(otherLabels)%>%
  left_join(labels)%>%
  mutate(labelStatus=ifelse(mwDist==0,'Same',ifelse(mwDist==1,'OneAWay','Other')))%>%
  mutate(ImageStatus=Image==OtherImage)%>%
  mutate(Image=as.factor(Image))


ggplot(subset(orig.cor,ImageStatus),aes(x=mwDist,y=corVal,col=Image))+
geom_jitter()+
geom_boxplot()+ggtitle('Uncorrected pixel correlations')

```


It doesn't look very different? We can try to calculate means of all the data.


```{r correlation values per image}

norm.cor%>%subset(ImageStatus)%>%
  group_by(Image,mwDist)%>%summarize(meanCor=mean(corVal))%>%
  ggplot(aes(x=mwDist,y=meanCor,col=Image))+geom_jitter()

norm.cor%>%subset(ImageStatus)%>%
  group_by(Image,labelStatus)%>%summarize(meanCor=mean(corVal))%>%
  ggplot(aes(x=labelStatus,y=meanCor,col=Image))+geom_jitter()


orig.cor%>%subset(ImageStatus)%>%
  group_by(Image,mwDist)%>%summarize(meanCor=mean(corVal))%>%
  ggplot(aes(x=mwDist,y=meanCor,col=Image))+geom_jitter()

orig.cor%>%subset(ImageStatus)%>%
  group_by(Image,labelStatus)%>%summarize(meanCor=mean(corVal))%>%
  ggplot(aes(x=labelStatus,y=meanCor,col=Image))+geom_jitter()

```


So this isnt' really behaving as expected? Or is it?