Analysing phenotype microarray data: `MongoDB` I/O

MongoDB is a popular document-oriented database. The RMongo package can be used in conjunction with opm to store and receive phenotype microarray data in such a database. OPMX objects fit nicely to such a pattern because they do not impose a certain structure on the metadata, much like MongoDB is able to store any kinds of data structures. The same holds for the options parts of the aggr_settings and disc_settings entries of OPMX objects. So the only work that needs to be done is to convert between OPMX objects and JSON strings.

See the documentation of MongoDB and RMongo for further details.

Author: Markus Goeker

library(RMongo)

## Loading required package: rJava

library(opm)

Define helper function

In our case, RMongo unfortunately returns a data frame with one plate per row and raw JSON strings per field. But the following short function is able to convert all such results. Note that opms also takes care of names converted by to_yaml, if any (see below).

mongo2opm <- function(x) {
  x <- split(x, seq_len(nrow(x)))
  x <- rapply(x, rjson::fromJSON, "character", NULL, "replace")
  opms(x, precomputed = FALSE, skip = FALSE, group = TRUE)
}

Create `MongoDB` connection

Connect to database test, which comes with MongoDB. We will use the collection pmdata to store JSON representations of OPM objects.

conn <- mongoDbConnect("test", "localhost", 27017)
coll <- "pmdata"

Insert plates

We insert each plate separately to be able to query them separately. Note that under these settings, to_yaml takes care of removing names with dots, which are disallowed in MongoDB.

result <- vapply(plates(vaas_4), function(plate) {
  dbInsertDocument(conn, coll, to_yaml(plate, json = TRUE, nodots = TRUE))
}, "")
stopifnot(result == "ok")
print(result)

## [1] "ok" "ok" "ok" "ok"

Search for plates with E. coli as species entry in the metadata

got <- dbGetQuery(conn, coll, '{"metadata.Species": "Escherichia coli"}')
stopifnot(is.data.frame(got), dim(got) > 0)

Convert to list of `OPMX` objects

Conversion necessary. See comments to mongo2opm above.

got <- mongo2opm(got)

Yields list with one element per plate type (only one plate type here). Some checks:

stopifnot(is.list(got), length(got) == 1, names(got) == plate_type(vaas_4))
got <- got[[1]]
stopifnot(is(got, "OPMS"), has_disc(got), dim(got) == c(2, dim(vaas_4)[-1]))
print(got)

## 1
## Class                 OPMD
## From file             ./E. coli DSM
##                       18039_vim10_12B__1_28_PMX_0_8#30#2010_E_12B_5.csv
## Hours measured        95.75
## Number of wells       96
## Plate type            Gen III
## Position              12-B
## Setup time            8/30/2010 1:19:11 PM
## Metadata              5
## Aggregated            TRUE
## Discretized           TRUE
## 
## 2
## Class                 OPMD
## From file             ./E. coli DSM
##                       30083T_vim10_7B__1_28_PMX_0_8#30#2010_F_
##                       7B_5.csv
## Hours measured        95.75
## Number of wells       96
## Plate type            Gen III
## Position              7-B
## Setup time            8/30/2010 1:53:08 PM
## Metadata              5
## Aggregated            TRUE
## Discretized           TRUE
## 
## => OPMS object with 2 plates (2 aggregated, 2 discretized) of type 'Gen III', 96 well(s) and about 384 time point(s).

Remove plates again

result <- dbRemoveQuery(conn, coll,
  '{$and: [{"csv_data": {$exists: true}}, {"measurements": {$exists: true}}]}')
stopifnot(result == "ok")
print(result)

## [1] "ok"

Check whether some plates are left over

empty <- dbGetQuery(conn, coll, '{"metadata.Species": "Escherichia coli"}')
stopifnot(is.data.frame(empty), dim(empty) == 0)

Close connection

dbDisconnect(conn)
rm(conn)

Tidy up

detach("package:RMongo", unload = TRUE)

Analysing phenotype microarray data: MongoDB I/O