05_clinical_variants_uniprot.Rmd

---
title: "clinical variants uniprot"
output: html_document
date: "2024-03-01"
---

```{r load and merge clinical variants}

library(ggplot2)
library(data.table)
library(stringr)
library(viridis)

base_dir=""
setwd(base_dir)

ranked_domains<-fread("analysis_files/domain_QC_summary_reproducibility_ranked.txt")

#load mutated domainome
mutated_domainome<-fread("analysis_files/mutated_domainome_merged_filtered.txt")

#load variant data
variant_data_uniprot<-fread("analysis_files/clinical_variants_uniprot_proteinAPI_domainomegenes.txt")
colnames(variant_data_uniprot)<-c("uniprot_ID","uniprot_entry","gene_name","sequence","type","begin","end","wt_aa","mut_aa","genomic_location","codon","consequence","source","str(somatic_status)","clinvar_id","exac_id","topmed_id","gnomad_id","NCITCGA_id","NCITCGA_cosmic_id","clingen_id","dbSNP_id","ensembl_id","str(sift_score)","sift_class","str(polyphen_score)","polyphen_class","clinvar_class","ensembl_class","uniprot_class","NCITCGA_class","descr")

mutated_domainome[, c("uniprot_ID") := tstrsplit(dom_ID, "_", fixed = TRUE)[1]]

mutated_domainome_variants<-merge(mutated_domainome,variant_data_uniprot,by.x=c("uniprot_ID","pos_in_uniprot","wt_aa","mut_aa"), by.y=c("uniprot_ID","begin","wt_aa","mut_aa"))
mutated_domainome_variants_all<-merge(mutated_domainome,variant_data_uniprot,by.x=c("uniprot_ID","pos_in_uniprot","wt_aa","mut_aa"), by.y=c("uniprot_ID","begin","wt_aa","mut_aa"),all.x = TRUE)

mutated_domainome_variants<-mutated_domainome_variants[!duplicated(aa_seq),]
mutated_domainome_variants_all<-mutated_domainome_variants_all[!duplicated(aa_seq),]

table(mutated_domainome_variants$clinvar_class)
table(mutated_domainome_variants$ensembl_class)
table(mutated_domainome_variants$sift_class)
table(mutated_domainome_variants$polyphen_class)


ggplot(mutated_domainome_variants[consequence=="missense" & !(clinvar_class==""),])+
  geom_density(aes(x=scaled_gr,col=clinvar_class))
ggplot(mutated_domainome_variants[consequence=="missense" & !(ensembl_class==""),])+
  geom_density(aes(x=scaled_gr,col=ensembl_class))

ggplot(mutated_domainome_variants[consequence=="missense" & !(sift_class==""),])+
  geom_density(aes(x=scaled_gr,col=sift_class))
ggplot(mutated_domainome_variants[consequence=="missense" & !(polyphen_class==""),])+
  geom_density(aes(x=scaled_gr,col=polyphen_class))


mutated_domainome_variants[clinvar_class %in% c("Pathogenic","Likely pathogenic","Risk factor"),clinvar:="Pathogenic"]
mutated_domainome_variants[clinvar_class %in% c("Benign","Likely benign"),clinvar:="Benign"]
mutated_domainome_variants[clinvar_class %in% c("Conflicting interpretations of pathogenicity"),clinvar:="Conflicting"]
mutated_domainome_variants[clinvar_class %in% c("Variant of uncertain significance"),clinvar:="Uncertain"]
mutated_domainome_variants[clinvar_class %in% c(""),clinvar:="no annotation"]

mutated_domainome_variants[ensembl_class %in% c("Pathogenic","Likely pathogenic","Risk factor"),ensembl:="Pathogenic"]
mutated_domainome_variants[ensembl_class %in% c("Benign","Likely benign"),ensembl:="Benign"]
mutated_domainome_variants[ensembl_class %in% c("Variant of uncertain significance"),ensembl:="Uncertain"]
mutated_domainome_variants[ensembl_class %in% c(""),ensembl:="no annotation"]

table(mutated_domainome_variants$clinvar,mutated_domainome_variants$ensembl)

mutated_domainome_variants[clinvar=="Pathogenic" | ensembl=="Pathogenic",clinical_class:="Pathogenic"]
mutated_domainome_variants[clinvar=="Benign" | ensembl=="Benign",clinical_class:="Benign"]
mutated_domainome_variants[clinvar=="Conflicting" | ensembl=="Conflicting",clinical_class:="Conflicting"]
mutated_domainome_variants[clinvar=="Uncertain" | ensembl=="Uncertain",clinical_class:="Uncertain"]
mutated_domainome_variants[clinvar=="no annotation" & ensembl=="no annotation",clinical_class:="no annotation"]


ggplot(mutated_domainome_variants[consequence=="missense",])+
  geom_density(aes(x=scaled_gr,col=clinvar))
ggplot(mutated_domainome_variants[consequence=="missense",])+
  geom_density(aes(x=scaled_gr,col=clinical_class))

table(mutated_domainome_variants[consequence=="missense",]$clinvar)
table(mutated_domainome_variants[consequence=="missense",]$clinical_class)

write.table(mutated_domainome_variants[consequence=="missense" & clinical_class != "no annotation",c("aa_seq","clinical_class")],file="analysis_files/clinical_variants_measured_uniprot_proteinAPI_labels.txt")


````