Usage

Finding papers in Pubmed

Reece et al (2010, doi:10.1016/j.ympev.2010.07.013) presented a phylogeny of moray eels using four different genes, but didn't publish the gene trees. I want to get the sequences underlying their analyses, which will be in the NCBI's databases, so I can reproduce their results. To get data associated with this paper from the NCBI I need the PMID (pubmed ID), which I can find using the rentrez function entrez_search to query the pubmed database with the paper's doi:

library("rentrez")
library("XML")

pubmed_search <- entrez_search(db = "pubmed", term = "10.1016/j.ympev.2010.07.013[doi]")
pubmed_search$ids

[1] "20674752"

All the functions in rentrez create a URL to get data from the NCBI, then fetch the resulting document, usually as an XML file. In most cases the functions will parse the most relevant sections of the XML file out and present them to you as items in a list (ids being one item of the pubmed_search list in this case).

Find data related to a particular record in an NCBI database

OK, now we have the PMID, what data does NCBI have for this paper? The entrez_link function lets us find out. In this case the db argument can be used to limit the number of data sources to check ( a list of avaliable linked-databases can be found via entrez_db_links(). In this case, I want to see every data source so I'll set db to “all”:

NCBI_data <- entrez_link(dbfrom = "pubmed", id = pubmed_search$ids, db = "all")
NCBI_data

elink result with ids from 13 databases:
 [1] pubmed_medgen              pubmed_nuccore            
 [3] pubmed_nucleotide          pubmed_pmc_refs           
 [5] pubmed_popset              pubmed_protein            
 [7] pubmed_pubmed              pubmed_pubmed_citedin     
 [9] pubmed_pubmed_combined     pubmed_pubmed_five        
[11] pubmed_pubmed_reviews      pubmed_pubmed_reviews_five
[13] pubmed_taxonomy_entrez    

Fetch a summary of a record

The most relevant data here is the from the popset database, which containts population and phylogenetic datasets. We can find out a little bit more about this datbase with entrez_db_summary()

entrez_db_summary("popset")

##                   DbName                 MenuName              Description 
##                 "popset"                 "PopSet" "PopSet sequence record" 
##                  DbBuild                    Count               LastUpdate 
##    "Build150625-0902m.1"                 "227554"       "2015/06/25 11:42"

If I want to see what each of the four popset datasets associated with this paper are about I can use entrez_summary to have a look. This function collects summaries of database records, and returns a list of parsed records with each element of a given record represented as the most natural base R type. In this case we can get summaries for each popset ID, check out the first record to see what kind of information they contain then extract the Title from each record:

data_summaries <- entrez_summary(db = "popset", id = NCBI_data$pubmed_popset)
data_summaries[[1]]

esummary result with 17 items:
 [1] uid        caption    title      extra      gi         settype   
 [7] createdate updatedate flags      taxid      authors    article   
[13] journal    strain     statistics properties oslt      

sapply(data_summaries, "[[", "title")

                                                                  307082412 
"Muraenidae cytochrome oxidase subunit 1 gene, partial cds; mitochondrial." 
                                                                  307075396 
         "Muraenidae recombination activating protein 2 gene, partial cds." 
                                                                  307075338 
         "Muraenidae recombination activating protein 1 gene, partial cds." 
                                                                  307075274 
                "Muraenidae cytochrome b gene, partial cds; mitochondrial." 

Fetch data

Ok, since we might expect nuclear and mitochondrial genes to hav different histories, let's get sequences from each genome (the the COI and RAG1 datasets) using entrez_fetch. By specifying rettype="fasta" we will get characater vectors in the fasta format:

coi <- entrez_fetch(db = "popset", rettype = 'fasta', id = NCBI_data$pubmed_popset[1])
rag1 <- entrez_fetch(db = "popset", rettype = 'fasta', id = NCBI_data$pubmed_popset[3])

So I've got the data on hand - that's all the I need rentrez for, but I might as well align these sequences and make gene trees for each. I'll just do a quick and diry neighbor-joining tree using ape and we can clean up the long OTU names with the help of stingr. (I put the fussy work of cleaning the names and rooting the trees into a function clean_and_root):

Note: This code block requires the user have have muscle installed on their $PATH, since many users won't hae this executable installed this code is not executed as part of the vignette

library(ape)
library(stringr)
clean_and_root <- function(tr, outgroup, resolved = TRUE) {

    tr$tip.label <- sapply(str_split(tr$tip.label, " "), function(x) paste(x[2:3],
        collapse = "_"))
    return(root(tr, outgroup, resolve.root = resolved))
}
write(coi, "~/moray_coi_raw.fasta")
write(rag1, "~/moray_rag1_raw.fasta")

par(mfrow = c(1, 2))

coi_ali <- muscle(read.dna("~/moray_coi_raw.fasta", "fasta"))
coi_tr <- nj(dist.dna(coi_ali, "k81"))
clean_coi_tr <- clean_and_root(coi_tr, "Uropterygius_macrocephalus")
plot(clean_coi_tr, direction = "rightwards", cex = 1)

rag_ali <- muscle(read.dna("~/moray_rag1_raw.fasta", "fasta"))
rag_tr <- nj(dist.dna(rag_ali, "k81"))
clean_rag_tr <- clean_and_root(rag_tr, "Uropterygius_macrocephalus")
plot(clean_rag_tr, direction = "leftward", cex = 1)

Working with WebEnvs

The NCBI provides search history features, which can be useful for dealing with a large lists of IDs (which will not fit in a single URL) or repeated searches. As an example, we will go searching for COI sequences from all the snail (Gastrppod) species we can find in the nucleotide database.

entrez_db_searchable helps us find fields that can make part of a search term.

search_fields <- entrez_db_searchable("nuccore")
search_fields

## Searchable fields for database 'nuccore'
##  [1] ALL  UID  FILT WORD TITL KYWD AUTH JOUR VOL  ISS  PAGE ORGN ACCN PACC
## [15] GENE PROT ECNO PDAT MDAT SUBS PROP SQID GPRJ SLEN FKEY PORG COMP ASSM
## [29] DIV  STRN ISOL CULT BRD  BIOS

search_fields$ORGN

## $Name
## [1] "ORGN"
## 
## $FullName
## [1] "Organism"
## 
## $Description
## [1] "Scientific and common names of organism, and all higher levels of taxonomy"
## 
## $TermCount
## [1] "3025070"
## 
## $IsDate
## [1] "N"
## 
## $IsNumerical
## [1] "N"
## 
## $SingleToken
## [1] "Y"
## 
## $Hierarchy
## [1] "Y"
## 
## $IsHidden
## [1] "N"

We can then use those terms in a call to entrez_search:

snail_search <- entrez_search(db = "nuccore", "Gastropoda[ORGN] AND COI[Gene]", usehistory = "y")

Because we set usehistory to “y” the snail_search object contains a unique ID for the search (WebEnv) and the particular query in that search history (QueryKey). Instead of using the hundreds of ids we turned up to make a new URL and fetch the sequences we can use the webhistory features.

cookie <- snail_search$WebEnv
qk <- snail_search$QueryKey
snail_coi <- entrez_fetch(db = "nuccore", WebEnv = cookie, query_key = qk, rettype = "fasta", retmax = 10)

In that case we used retmax to limit the number of queries we downloaded. There are actually thousands of records. If we wanted to download all of them it would probably be a good idea to downlaod them in batches (both to give the NCBI's severs a break and to make sure a corrupted download doesn't ruin your whole process. Using a for loop in conjunction with the terms restart and retmax we can download the sequences 50 at a time. (This code is not run as part of this vignette so save you time and the NCBI some server-load):

for (start_rec in seq(0, 200, 50)) {
    fname <- paste("snail_coi_", start_rec, ".fasta", sep = "")
    recs <- entrez_fetch(db = "nuccore", WebEnv = cookie, query_key = qk, rettype = "fasta", retstart = start_rec, retmax = 50)
    write(recs, fname)
    print(paste("wrote records to ", fname, sep = ""))
}