map_peptides.R

## MAP AAS TO ENSEMBL PROTEINS & TRANSCRIPTS
## and collect sequence & structure data at AAS.


library(segmenTools)
options(stringsAsFactors=FALSE)


## DATA FROM  genomeBrowser, project folder data/mammary,
## run steps in data/mammary/setup.sh to create all data
## required here!

mam.path <- file.path(Sys.getenv("MAMDATA")) 
if ( mam.path=="" ) # author's local path
    mam.path <- "/home/raim/data/mammary"

feature.file <- file.path(mam.path,"features_GRCh38.110.tsv")

if ( !file.exists(feature.file) )
    stop("genome feature table file not found. This script requires ",
         "setup of genomic data via the genomeBrowser/data/mammary/setup.sh. ",
         "If you have set this up, please provide the path here as `mam.path` ",
         "and either change the path of the saap_mapped.tsv input or gzip and ",
         "copy it from processedData to additionalData.\n",
         "NOTE, that the downstream analysis can still be run, since we ",
         "provide the output of this script, saap_mapped.tsv")


## DECODE DATA
proj.path <- file.path(Sys.getenv("DECODE")) 
if ( proj.path=="" ) # author's local path
    proj.path <- "/home/raim/data/decode"

dat.path <- file.path(proj.path,"originalData")
fig.path <- file.path(proj.path,"figures","saap_mapping")
out.path <- file.path(proj.path,"processedData")
out.file <- file.path(out.path,"saap_mapped.tsv")

## SET THIS VARIABLE TO GENERATE PDFs INSTEAD OF PNGs
ftyp <- "png"
##if ( !interactive() ) ftyp <- "pdf"

## SAAP/BP pairs
saapf <- file.path(out.path,"unique_saap.tsv")
## BP: protein blast results
bpmap <- file.path(out.path,"bp_mapped.tsv")

## protein fasta (Ensembl+Mutations, used for blast)
pfasta <- file.path(out.path,"all_proteins.fa")

## protein:transcript ID mapping
tpmap.file <- file.path(mam.path,"originalData","protein_transcript_map.tsv")

## coding region fasta
tfasta <- file.path(mam.path,"processedData","coding.fa")

## UTR Lengths
utr.file <- file.path(mam.path, "processedData", "transcript_utr_lengths.tsv")

## protein-transcript map
tpmap <- file.path(mam.path,"originalData","protein_transcript_map.tsv")
## CDS structure
cdsmap <- file.path(mam.path,"processedData","protein_cds_structure.dat")
cdspos <- file.path(mam.path,"processedData","protein_cds_coordinates.tsv")


## structure predictions
## s4pred
s4pred <- file.path(mam.path,"processedData",
                    "Homo_sapiens.GRCh38.pep.large_s4pred.fas.gz")
## iupred3/anchor2
iupred <- file.path(mam.path,"processedData","iupred3")

## uniprot<->ensembl mapping: required to find describePROT
uni2ens.file <- file.path(mam.path,"originalData","uniprot_ensembl.dat")

## describePROT
##  * sequence,
##  * MMseqs2 - Fast sequence alignment [PMID:30615063],
##  * ASAquick - Prediction of protein accessible surface area [PMID:27787824],
##  * DisoRDPbind - Prediction of disordered RNA, DNA, and protein
##    binding residues [PMID:26109352],
##  * SCRIBER - Prediction of protein binding residues [PMID:31510679],
##  * flDPnn - Prediction of intrinsically disordered residues [PMID:34290238].
descrp <- file.path(mam.path,"processedData","describePROT")

## pfam hits
pfam.file <- file.path(mam.path,"processedData",
                  "Homo_sapiens.GRCh38.pep.large_annotations.csv")
## pfam download from EBI
pfdl.file <- file.path(mam.path,"originalData", "9606.tsv.gz")
## pfam clans
clans.file <- file.path(mam.path,"originalData", "pfam", "Pfam-A.clans.tsv.gz")

## copied from the Biostrings package
GENETIC_CODE <- c(
    TTT="F", TTC="F",         
    TTA="L", TTG="L",         
    TCT="S", TCC="S", TCA="S", TCG="S", AGT="S", AGC="S",         
    TAT="Y", TAC="Y",         
    TAA="*", TAG="*", TGA="*",
    TGT="C", TGC="C",    
    TGG="W",         
    CTT="L", CTC="L", CTA="L", CTG="L",         
    CCT="P", CCC="P", CCA="P", CCG="P",         
    CAT="H", CAC="H",         
    CAA="Q", CAG="Q",         
    CGT="R", CGC="R", CGA="R", CGG="R", AGA="R", AGG="R",         
    ATT="I", ATC="I", ATA="I",
    ATG="M",         
    ACT="T", ACC="T", ACA="T", ACG="T",         
    AAT="N", AAC="N",         
    AAA="K", AAG="K",         
    GTT="V", GTC="V", GTA="V", GTG="V",         
    GCT="A", GCC="A", GCA="A", GCG="A",         
    GAT="D", GAC="D",         
    GAA="E", GAG="E",         
    GGT="G", GGC="G", GGA="G", GGG="G"
)

## figures
dir.create(fig.path, showWarnings=FALSE)



#### FIND POSITIONS of SAAP in protein and transcript

## GET ENSEMBL PROTEINS - from project mammary
fas <- readFASTA(pfasta, grepID=TRUE)

## get matching transcripts
trfas <- readFASTA(tfasta, grepID=TRUE)

## protein-transcript map
trmap <- read.delim(file=tpmap, header=FALSE, row.names=2)
## reverse map transcript-protein
pamrt <- matrix(rownames(trmap), ncol=1)
rownames(pamrt) <- trmap[,1]

## rename by protein names via trmap, for quick access
## of protein-specific transcript
names(trfas) <- pamrt[names(trfas),1]


## list of splice sites for each CDS
cds <- read.delim(cdsmap,header=FALSE, row.names=1)
cdl <- strsplit(cds[,1],";")
names(cdl) <- rownames(cds)
cdl <- lapply(cdl, as.numeric)
## genomic position of CDS
cpos <- read.delim(cdspos, header=FALSE, sep=" ")
colnames(cpos) <- c("ID","chr","start","end","strand")
posl <- split(cpos[,2:5], cpos[,1])
## NOTE: reverse order for minus strand
posl <- lapply(posl, function(x) x[order(x$start,
                                         decreasing=x$strand=="-"),])
## transcript positions: add UTRs
utr <- read.delim(utr.file, row.names=1)
rownames(utr) <- pamrt[rownames(utr),1]


## s4pred
s4p <- readFASTA(s4pred, grepID=TRUE)
## skip version number - TODO: use version numbers?
names(s4p) <- sub("\\.[0-9]+", "", names(s4p))
## reorder
cat(paste("s4pred:", sum(!names(fas)%in%names(s4p)), "proteins not found\n"))
s4p <- s4p[names(fas)]

## map iupred files here! 
iufiles <- list.files(pattern=paste0(".*iupred3.tsv.gz"), path=iupred)
names(iufiles) <- sub("\\..*","", iufiles)
iufiles <- iufiles[names(fas)]

cat(paste("iupred3", sum(is.na(iufiles)), "proteins not found\n"))

## read uniprot<->ensembl mapping for describPROT
uni2ens <- read.delim(uni2ens.file, header=FALSE)
uni2ens[,2] <- sub("\\..*", "", uni2ens[,2]) # remove ensembl version tag
uni2ens[,1] <- sub("-.*", "", uni2ens[,1]) # remove uniprot version tag
uni2ens <- uni2ens[uni2ens[,2]%in%names(fas),]

## PFAM 37.0 ANNOTATION from EBI
pfd <- read.delim(pfdl.file, skip=3, header=FALSE)
colnames(pfd) <- c("seq id", "alignment start",
                   "alignment end",
                   "FROM",#"envelope start",
                   "TO",#"envelope end",
                   "hmm acc", "hmm name", "type",
                   "hmm start", "hmm end", "hmm length",
                   "bit score", "E-value", "clan")
## add ensembl ID
pfd$ensembl <- uni2ens[match(pfd[,1], uni2ens[,1]),2]

## OWN PFAM/hmmer predictions 
pfm <- read.csv(file=pfam.file, sep=";", fill=FALSE,
                header=FALSE, comment.char="#")
pfmh <- c(
    "target", "tid" , "tlen",            
    "query"   , "qid" , "qlen",                 
    "E-value"      , "score"     , "bias",                 
    "#"            , "of"        , "c-Evalue",             
    "i-Evalue"     , "score"     , "bias",            
    "from"         , "to"        , "from",                 
    "to"           , "FROM"      , "TO",                   
    "acc")#          , "description")
colnames(pfm) <- pfmh

## pfam clans: used to collapse domains below!
pclan <- read.delim(clans.file, row.names=1 , header=FALSE)
pfm$clan <- pclan[sub("\\..*","", pfm$tid),"V2"]
## same tag as EBI PFAM for No_clan
pfm$clan[pfm$clan==""] <- "No_clan"

## pfam w/o version number
pfm$pfam <- sub("\\..*","", pfm$tid)

## remove version number for access
pfm$ensembl <- sub("\\..*","", pfm$query)


### READ IN and MERGE unique SAAP/BP pairs and BP blast results
dat <- read.delim(saapf, header=FALSE)
colnames(dat) <- c("SAAP","BP")
bmap <- read.delim(bpmap)

dat <- merge(dat, bmap, by="BP", all=TRUE)


## result vectors
mut <- pos <- len <- cdn <- tps <- aaf <- aat <-  aas <-
    sss <- anc <- iup <- iubg <- anbg <-
        mmseq2 <- asaquick <- disordRDPbind <- scriber <- flDPnn <-
            pctx <- nctx <- pfam <- clan <- pfam.ebi <- clan.ebi <-
                rep(NA, nrow(dat))

## secondary structure frequencies in whole protein
sssbg <- matrix(NA, nrow=nrow(dat), ncol=3)
colnames(sssbg) <- c("C","E","H")

## genomic coordinate of AAS
gcoor <- matrix(NA, nrow=nrow(dat), ncol=3)
colnames(gcoor) <- c("chr","coor","strand")

## Quality Control
testit <- TRUE # TEST whether the mutation position is correct
use.regex <- TRUE #FALSE # use TRUE to test blast results vs. direct regex

## count errors
errtypes <- c("no protein", "using blast","no BP",
              "wrong s4pred len","wrong iupred3 len",
              "no describePROT",
              "wrong descPROT",
              "no transcript", "wrong codon",
              "AAS > CDS len")
wiup <- character()
errors <- matrix(0, nrow=nrow(dat), ncol=length(errtypes))
colnames(errors) <- errtypes
if ( !use.regex ) errors[,"using blast"] <- 1


## trouble shooting 20240424
pid="ENSP00000354876" # no transcript even though its in protein_transcript_map
## it's not in coding.fa, why??

## 20240807 #EGLELLK")
i=which(dat$BP=="ACQRPQLWQTIQTQGHFQLQLPPGK" &
        dat$SAAP=="ACQQPQLWQTIQTQGHFQLQLPPGK")

## 2002408

for ( i in 1:nrow(dat) ) {
   
    oid <- dat$protein[i] # ID with mutation index
    gid <- dat$ensembl[i] # original gene ID

    j <- which(names(fas)==oid)
    if ( length(j)==0 ) { # 
        cat(paste("WARNING:",i, oid, "no protein sequence found\n"))
        errors[i,"no protein"] <- 1
        next
    }
    if ( length(j)>1 ) { # atm not appearing
        stop(paste("WARNING:",i, oid, "more than one hit\n"))
        next
    }
    query <- unlist(dat[i,"BP"])
    target <- fas[[j]]$seq

    ## GET POSITION OF MUTATION
    ## 1: MAP PETPTIDE and test consistency with blast
    if ( use.regex ) {
        res <- gregexpr(query, target)
        if ( length(res)>1 ) {
            stop(paste("PROBLEM:", i, oid, 
                       "more than two hits in",j, names(fas)[j], "\n"))
        } else if ( res[[1]][1]==-1 ) {
            cat(paste("WARNING:",i, oid, "no match in", j, names(fas)[j],
                      ", using blast result with",
                      dat$mismatches[i]," mismatches, class:",dat$match[i],"\n"))
            errors[i,"using blast"] <- 1
            AAS <- dat[i,"sstart"]
        } else if ( res[[1]][1] != dat[i,"sstart"] ) {
            ## if this never occurs we could use blast results!
            stop(paste("PROBLEM:", i, oid, 
                       "regex doesn't match blast\n"))
        } else  AAS <- res[[1]][1]
    } else AAS <- dat[i,"sstart"]

    ## 2: AAS within protein
    saap <- unlist(dat[i,"SAAP"])
    mut[i] <- which(strsplit(saap,"")[[1]]!=strsplit(query,"")[[1]])
    pos[i] <- AAS + mut[i] -1  ## POSITION IN PROTEIN!
    len[i] <- length(unlist(strsplit(target,"")))

    ## GET AA SEQUENCE CONTEXT +/-25
    DST <- 25
    nsq <- nchar(target)
    rrng <- seq(-DST,DST,1)
    sq <- rep("-",length(rrng)) ## GAP
    names(sq) <- rrng
    ## GET RANGE AROUND AAS
    rng <- (pos[i]-DST):(pos[i]+DST)
    ## cut range to available
    rrng <- rrng[rng>0 & rng<=nsq]
    rng <- rng[rng>0 & rng<=nsq]
    sq[as.character(rrng)] <- unlist(strsplit(target,""))[rng]
    pctx[i] <- paste0(sq, collapse="")
   
    ## test mutation
    if ( testit ) { #
        if ( length(grep(query,target))==0 ) {
            cat(paste("WARNING:", i, gid, j, names(fas)[j],
                      "BP is not in target protein\n"))
            errors[i,"no BP"] <- 1
        } else {
            ntarg <- sub(query,saap,target)
            if ( strsplit(target,"")[[1]][pos[i]] ==
                 strsplit(ntarg,"")[[1]][pos[i]] ) {
                stop(paste("WARNING:", i, gid, j, names(fas)[j],
                           "no mutation detected\n"))
                
            }
        }
    }

    ## get PFAM CLANS - EBI download
    pfams <- pfd[which(pfd$ensembl==gid),,drop=FALSE]
    if ( nrow(pfams)> 0) {
        pidx <- which(pfams$FROM <= pos[i] & pfams$TO >= pos[i])
        if ( length(pidx)>0 ) {
            clan.ebi[i] <- paste0(unique(pfams[pidx, "clan"]), collapse=";")
            pfam.ebi[i] <- paste0(unique(pfams[pidx, "hmm acc"]), collapse=";")
        }
    }
    ## get PFAM CLANS - own hmmer prediction
    pfams <- pfm[which(pfm$ensembl==gid),,drop=FALSE]
    if ( nrow(pfams)> 0) {
        pidx <- which(pfams$FROM <= pos[i] & pfams$TO >= pos[i])
        if ( length(pidx)>0 ) {
            clan[i] <- paste0(unique(pfams[pidx, "clan"]), collapse=";")
            pfam[i] <- paste0(unique(pfams[pidx, "tid"]), collapse=";")
        }
    }

    ## GET S4Pred
    s4s <- s4p[[gid]]
    if ( is.null(s4s) ) {
        cat(paste("WARNING:", i, "no s4pred found for", oid, "\n"))
        errors[,"wrong s4pred len"] <- "no s4pred result"
    } else {
        ## test length
        if ( nchar(s4s$seq)!=len[i] ) {
            cat(paste("WARNING:", i, "s4 prediction has wrong length for",
                      oid, "\n"))
            errors[,"wrong s4pred len"] <- 1
        } else {
            sss[i] <- substr(s4s$seq, pos[i], pos[i])
            ## background: complete protein sequence
            tb <- table(unlist(strsplit(s4s$seq,"")))
            sssbg[i,names(tb)] <- tb
        }
    }

    ## GET IUPred3
    ## TODO: remove version tag from iupred files to avoid searching!
    ## or search once above
    iufile <- iufiles[gid]
    if ( is.na(iufile) )  {
        cat(paste("WARNING:", i, "no iupred3 found for", oid, "\n"))
        wiup <- c(wiup, paste(i, gid, "not found"))
        errors[i,"wrong iupred3 len"] <- "no iupred3 result"
    } else {
        iud <- read.delim(file.path(iupred,iufile), header=FALSE)
        if ( nrow(iud)!=len[i] ) {
            cat(paste("WARNING:", i, "iupred3 prediction has wrong length for",
                      oid, "iupred:",nrow(iud), "vs. protein:", len[i],"\n"))
            errors[i,"wrong iupred3 len"] <- 1
        } else {
            anc[i] <- iud[pos[i], 4] #anchor2
            iup[i] <- iud[pos[i], 3] #iupred3
            ## whole protein mean
            anbg[i] <- mean(iud[, 4])
            iubg[i] <- mean(iud[, 3])
        }
    }

    ## get describePROT data
    if ( gid%in%uni2ens[,2] ) {
        idx <- which(uni2ens[,2]==gid)
        if ( length(idx)>1 ) {
            cat(paste("WARNING:", i, gid, "multiple uniprot hits,",
                      "taking first\n"))
            idx <- idx[1]
        }
        uid <- uni2ens[idx,1]
        ## open file
        dprt.file <- file.path(descrp, paste0(uid,".tsv.gz"))
        if ( !file.exists(dprt.file) ) {
                cat(paste("WARNING:", i,
                          "desribePROT missing for",
                          oid, uid, "\n"))
                errors[i,"no describePROT"] <- 1
        } else {
            dprt <- read.delim(dprt.file, header=FALSE)

            ## check sequence
            if ( nrow(dprt)!=len[i] ) {
                cat(paste("WARNING:", i,
                          "desribePROT has wrong length for",
                          oid, uid, "desribeP:",nrow(dprt), "vs. protein:",
                          len[i],"\n"))
                errors[i,"wrong descPROT"] <- 1
            }             
            mmseq2[i] <- dprt[pos[i], 2] 
            asaquick[i] <- dprt[pos[i], 3] 
            disordRDPbind[i] <- dprt[pos[i], 4] 
            scriber[i] <- dprt[pos[i], 5] 
            flDPnn[i] <- dprt[pos[i], 6] 
        }
    }
    
    ## GET CODON
    ## protein:transcript mapping
    ## load coding region get codon
    nt <- trfas[[gid]]
    if ( is.null(nt) ) {
        cat(paste("WARNING:", i, "no transcript found for", oid, "\n"))
        errors[i,"no transcript"] <- 1
        next
    }

    ## position of first codon position
    npos <- (pos[i]-1)*3+1
    ## retrieve codon from the transcript sequence
    codon <- substr(nt$seq, npos, npos+2)

    ## STORE FROM/TO AA
    aas[i] <- strsplit(target,"")[[1]][pos[i]]
    aaf[i] <- strsplit(query,"")[[1]][mut[i]]
    aat[i] <- strsplit(saap,"")[[1]][mut[i]]

    ## STORE CODON
    ## don't store wrong codon!
    if ( GENETIC_CODE[codon] != aaf[i] ) {
        cat(paste("WARNING:",i,"wrong codon", aaf[i],
                  "vs", codon, GENETIC_CODE[codon],"in", oid, "\n"))
        errors[i,"wrong codon"] <- 1
    } else {
        ## store codon
        cdn[i] <- codon
        ## store position in transcript
        if ( !gid%in%rownames(utr) )
            stop("missing UTR data")
        tps[i] <- npos + utr[gid,"start"] +1 # 2nd codon pos
    }
    
    ## GET NT SEQUENCE CONTEXT +/-25
    fsq <- nt$seq # transcript sequence
    nsq <- nchar(fsq)
    
    rrng <- seq(-DST*3,DST*3 +2,1)
    sq <- rep("-",length(rrng)) ## GAP
    names(sq) <- rrng

    ## GET RANGE AROUND AAS
    tpos <- pos[i]*3 -2
    rng <- (tpos-DST*3):(tpos+DST*3 +2)
    ## cut range to available
    rrng <- rrng[rng>0 & rng<=nsq]
    rng <- rng[rng>0 & rng<=nsq]
    sq[as.character(rrng)] <- unlist(strsplit(fsq,""))[rng]
    nctx[i] <- paste0(sq, collapse="")

    ## GET GENOMIC POSITION via CDS list and posl
    cds <- cdl[[gid]] # position of splice sites in transcript
    coors <- posl[[gid]] # genomic positions of CDS

    ## get second codon position
    npos <- npos+1

    exon <- which(cds>=npos)[1] # in which exon is the 2nd codon position
    if ( is.na(exon) ) {
        cat(paste("WARNING:",i,"wrong genomic position", npos, "vs", max(cds),
                  "in", gid, "\n"))
         errors[i,"AAS > CDS len"] <- 1
    } else {
    
        rpos <- npos # relative position in exon/CDS
        if ( exon>1 ) # subtract prior exons/CDS
            rpos <- npos-cds[exon-1]

        ## get genomic position of AAS 2nd codon position
        chr <- unique(coors$chr)
        strand <- unique(coors$strand)
        
        if ( strand=="-") {
            coor <- coors[exon, "end"] - rpos +1
        } else {
            coor <- coors[exon, "start"] + rpos -1
        }
        gcoor[i,] <- c(chr, coor, strand)
    }
     
    ## report
    if ( !interactive() ) {        
        tcodons <- paste(names(which(GENETIC_CODE%in%aat[i])),collapse=";")
        cat(paste("DONE:", i, codon, aaf[i], GENETIC_CODE[codon], "->", aat[i],
                  tcodons,  "\n"))
    }
    
    if ( aas[i] != aaf[i] ) 
        warning(paste("WARNING: BP pos from AA:", aaf[i], "vs",
                  aas[i], GENETIC_CODE[codon], gid, "\n"))
}






## relative position
rpos <- pos/len
## bind to data frame
bgsss <- sssbg
colnames(bgsss) <- paste0(colnames(bgsss), ".protein")
dat <- cbind(dat,
             site=mut,  # mutated site within peptide
             pos=pos,   # position of AAS within protein
             len=len,
             rpos=pos/len,
             from=aaf, to=aat,
             codon=cdn, 
             tpos=tps,   # position of AAS codon within transcript
             gcoor,      # genome coordinate of AAS
             s4pred=sss, bgsss, iupred3=iup, iupred3.protein=iubg,
             anchor2=anc, anchor2.protein=anbg,
             MMSeq2=mmseq2,
             ASAquick=asaquick,
             DisoRDPbind=disordRDPbind,
             SCRIBER=scriber,
             flDPnn=flDPnn,
             AA=pctx,
             NT=nctx,
             pfam=pfam,
             clan=clan,
             pfam.ebi=pfam.ebi,
             clan.ebi=clan.ebi)

## FILTER DATA

## missing position - no match found from peptide in protein!!
## TODO: perhaps test LONGEST of the leading razor proteins.
rm <- which(is.na(dat$pos)) #dat$pos<0)
if ( length(rm)>0 ) {
    ##dat <- dat[-rm,]
    ##errors <- errors[-rm,]
    cat(paste("WARNING:", length(rm), "SAAP without protein match\n"))
}

## remove where length or position are NA - same as "no match"!
rm <- which(is.na(dat$len))
if ( length(rm)>0 ) {
    ##dat <- dat[-rm,]
    ##errors <- errors[-rm,]
    cat(paste("WARNING:", length(rm), "SAAP no protein length\n"))
}

## remove where no codon was assigned
## no transcript found or wrong codon
rm <- which(is.na(dat$codon))
if ( length(rm)>0 ) {
    ##dat <- dat[-rm,]
    cat(paste("WARNING:", length(rm), "SAAP w/o codon\n"))
}

## remove where no codon was assigned
## no transcript found or wrong codon
rm <- which(is.na(dat$chr))
if ( length(rm)>0 ) {
    ##dat <- dat[-rm,]
    cat(paste("WARNING:", length(rm), "SAAP w/o genome coordinates\n"))
}

### ERROR AND RECOVERY STATS

## NOTE:
## * "no BP" and "using blast" are the same, and all are "bad" or "wrong"
##   blast matches,
## * TCHVEHPSLQNPITVEW:TCHVEHPSLQSPITVEW:ENSP00000382018
##   is classified as "wrong" in blast, because it is associated with
##   a scaffold gene (is.na(all$gene)) and not with the official genome gff3.

errs <- errors#[,apply(errors,2,sum)>0]
er <- c("unique BP/SAAP"=nrow(dat),
        "unique SAAP"=sum(!duplicated(dat$SAAP)),
        "unique BP"=sum(!duplicated(dat$BP)),
        table(dat$match)[c("good","bad","wrong")],
        "no gene"=sum(dat$nogene, na.rm=TRUE),
        "with mutation"=sum(dat[,"protein"]!=dat[,"ensembl"],na.rm=TRUE),
        apply(errs,2,sum))
plotdev(file.path(fig.path,"mapping_errors"),
        res=300, width=4, height=3.5, type=ftyp)
par(mai=c(1.5,.5,.5,.1), mgp=c(1.3,.3,0), tcl=-.25)
bp <- barplot(er,las=2)
text(bp+.25, er+.05*diff(par("usr")[3:4]),
     labels=er, pos=3, xpd=TRUE, cex=.8, srt=90)
dev.off()

## TODO: better characterize matches to mutated proteins
recovery <- c("SAAP"=nrow(dat),
              "protein"=sum(!is.na(dat[,"protein"])),
              ##"length"=sum(!is.na(dat[,"len"])),
              "s4pred"=sum(!is.na(dat[,"s4pred"])),
              "iupred3"=sum(!is.na(dat[,"iupred3"])),
              "genome"=sum(!is.na(dat[,"chr"])),
              "codon"=sum(!is.na(dat[,"codon"])))
plotdev(file.path(fig.path,"mapping_recovery"),
        res=300, width=4, height=3.5, type=ftyp)
par(mai=c(.75,.5,.25,.1), mgp=c(1.3,.3,0), tcl=-.25)
bp <- barplot(recovery,las=2,ylab="")
text(bp, recovery, recovery, pos=3, xpd=TRUE, cex=.8)
dev.off()

types <- c("SAAP"=nrow(dat),
           "MANE"=sum(!is.na(dat[,"MANE.protein"])),
           "extracellular"=sum(dat[,"extracellular"], na.rm=TRUE),
           "IG"=sum(dat[,"IG"], na.rm=TRUE),
           "albumin"=sum(dat[,"albumin"], na.rm=TRUE),
           "globin"=sum(dat[,"globin"], na.rm=TRUE))

types <- sort(types, decreasing=TRUE)
plotdev(file.path(fig.path,"mapping_types"),
        res=300, width=4, height=3.5, type=ftyp)
par(mai=c(.75,.5,.25,.1), mgp=c(1.3,.3,0), tcl=-.25, cex=.8)
bp <- barplot(types,las=2,ylab="")
text(bp, types, types, pos=3, xpd=TRUE)
dev.off()

## inspect some overlaps
table(dat[,"extracellular"], dat[,"globin"])
table(dat[,"extracellular"], dat[,"albumin"])
table(dat[,"extracellular"], dat[,"IG"])


### WRITE OUT TABLE with positions for downstream analysis
write.table(dat, file=out.file,
                sep="\t", quote=FALSE, na="", row.names=FALSE)