An introduction to opm opm – an R package to analyse ... · opm – an R package to analyse...

Leibniz-Institut DSMZ-Deutsche Sammlung von Mikroorganismen und Zellkulturen GmbH

opm – an R package to analyse OmniLog Phenotype MicroArray data

Dr. Johannes Sikorski, Dr. Lea Vaas, Dr. Markus GökerLeibniz-Institut DSMZ-Deutsche Sammlung von Mikroorganismen und Zellkulturen GmbHwww.dsmz.de

An introduction to opm


You have numerous OmniLog Phenotype MicroArray data

and you want to explore them full-fledged and quantitatively into diverse directions of analysis frameworks.

• of closely related organisms or cell lines

• of numerous well-defined mutants

• obtained under diverse physiological test conditions

www.biolog.com


to organize your PM data, curve parameters and metadata

to subset and query your data

graphical display of raw kinetics or aggregated curve parameters

exploit the full statistics implemented in R

export to third-party software using YAML

opm: Tools for analysing OmniLog(R) Phenotype Microarray data

enables you:


http://www.rstudio.org/RStudio™ is a free and open source integrated development environment (IDE) for R.

http://www.r-project.org/R is a free software environment for statistical computing and graphics.

http://cran.r-project.org/web/packages/opm/index.htmladd-on package opm: Tools for analysing OmniLog(R) Phenotype Microarray data

Software requirements

http://www.r-project.org/�


The R code of this presentation is available on request from

Dr. Johannes Sikorski [email protected]

R Code of this presentation

Dr. Lea Vaas [email protected]

Dr. Markus Göker [email protected]

Feel free to contact us in case of any questions regarding usage of opm.


opm enables you:

to organize your PM data, curve parameters and metadata

to subset and query your data

graphical display of raw kinetics or aggregated curve parameters

exploit the full statistics implemented in R

export to third-party software using YAML


OPMorganizes your PM data in OPMS objects:

Plate 1 Plate 2 Plate 3Plate 4 Plate 5 Plate 6

Plate 7 Plate 8 Plate 9





raw kinetic data

aggregated curve

parameters

metadata

An OPMS object stores:

Example: a set of 9 PM plates of the same plate type

The size of the OPMS object is only limited by the amount of RAM memory

Hour00.0000.2500.50.30.00.60.00

lysin353337.

102.

328Hourinte

nsity

per well:raw kinetic data

metadata Plate 3

Taxonomy Bacillus subtilis.habitat soilsampling place GPS coord.sampling date 2011-06-15sampling season summerhabitat [°C] 27.sporulation yes.PCR (gene xyz) positive.... as much and what you wish...

per plate:any metadata of interest to the user

lysin

mu 15.559078 lambda 5.798210A 305.989319AUC 23308.269348mu CI95 low 3.803466lambda CI95 low 1.080333A CI95 low 305.642353AUC CI95 low 23125.092442mu CI95 high 140.841704lambda CI95 high 11.819251A CI95 high 306.986123AUC CI95 high 23411.648024

per well:aggregated curve parameters, confidence-intervals from bootstrapping

Lag = lambda, Slope = mu, Max = A, Area Under the Curve = AUC


read_opm()


Load Demo files that come with the opm package

# Use the built-in function opm_files() to retrieve the paths where the example files in your R installation are located:(files <- opm_files("testdata"))

# read in the files, which are zipped# using the include argument to select specific plates of interest# by this, three files are loaded into the object "example.opm"example.opm <- read_opm(files, include = "*Example_?.csv.xz")

read_opm(names, convert = c("try", "no", "yes", "sep", "grp"), gen.iii = FALSE, include = list(), ..., demo = FALSE)


# read in all CSV raw data files in your working directoryPM1 <- read_opm(".")

# read in all CSV raw data files in your working directory and convert the plate type to GenIII platesGenIII <- read_opm(".", gen.iii = TRUE)


read_opm(names, convert = c("try", "no", "yes", "sep", "grp"), gen.iii = FALSE, include = list(), ..., demo = FALSE)



# let us check some information on the files in this OPMS objectplates(example.opm)summary(example.opm)show(example.opm)dim(example.opm)hours(example.opm)length(example.opm)max(example.opm)plate_type(example.opm)seq(example.opm)setup_time(example.opm)measurements(example.opm)wells(example.opm)wells(example.opm, full = TRUE)




returns the raw kinetic data


do_aggr()

x <- do_aggr(example.opm, program = "opm-fast")aggregate only “A“ and “AUC“ using a fast algorithm

x <- do_aggr(example.opm)aggregate all 4 parameters using a spline fit algorithm (grofit package)

x <- do_aggr(example.opm, program = "opm-fast", boot = 100)include 100x bootstrap replicates

x <- do_aggr(example.opm , boot = 100) Note: time consuming


aggregated(example.opm)check aggregated data

A01 A02 A03 A04 A07mu 4.242159 5.769109 0.02138581 0.2827407 0.2383062lambda -2.340620 12.799329 -465.46803431 20.0749555 -14.4573092A 47.923185 62.738943 11.51078807 19.4617762 18.2811191AUC 3914.852139 4154.830048 1070.20657323 1250.9426009 1396.9447154mu CI95 low 2.733574 3.045267 -1.10076311 -2.2050686 -4.8515830lambda CI95 low -38.403543 -10.300782 56.14216650 42.4248855 24.8184260A CI95 low 47.197513 58.940763 11.17285004 19.1992801 16.9627344AUC CI95 low 3875.243148 4093.577722 1056.62986435 1230.3571787 1352.9702303mu CI95 high 14.170557 13.689212 6.15737265 9.3063345 21.5309783lambda CI95 high 79.044830 50.248293 87.70587107 106.1197708 107.3697670A CI95 high 52.484756 67.456369 15.37628753 23.6590936 30.0717055AUC CI95 high 3941.361758 4183.239559 1077.02925382 1262.9208049 1432.5071603


OPMorganizes your PM data in OPMS objects:







raw kinetic data

aggregated curve

parameters

metadata

An OPMS object stores:

Example: a set of 9 PM plates of the same plate type

You need to provide the metadata

separately

Hour00.0000.2500.50.30.00.60.00

lysin353337.

102.

328Hourinte

nsity

per well:raw kinetic data

metadata Plate 3

Taxonomy Bacillus subtilis.habitat soilsampling place GPS coord.sampling date 2011-06-15sampling season summerhabitat [°C] 27.sporulation yes.PCR (gene xyz) positive.... as much and what you wish...

per plate:any metadata of interest to the user

lysin

mu 15.559078 lambda 5.798210A 305.989319AUC 23308.269348mu CI95 low 3.803466lambda CI95 low 1.080333A CI95 low 305.642353AUC CI95 low 23125.092442mu CI95 high 140.841704lambda CI95 high 11.819251A CI95 high 306.986123AUC CI95 high 23411.648024

per well:aggregated curve parameters, confidence-intervals from bootstrapping

Lag = lambda, Slope = mu, Max = A, Area Under the Curve = AUC


We use as identifier the Setup Time and Position of the plate in the reader.

Good news:opm allows to export these informations as a start for the metadata file using the function:collect_template()

You need to provide the metadata separately

One Problem Arises:Imagine, you have numerous plates with numerous metadata to each plate. How can we make sure that the metadata are matched CORRECTLY to the specific raw kinetic data?

Solution:We need an identifier that perfectly matches metadata to raw kinetic data.


collect_template()

data frame

add further metadata columns

metadata <- collect_template(files, include = "*Example_?.csv.xz")

Unique identifier to merge metadata and raw kinetic data


collect_template()

data frame


CSV file (or *.txt, *.dat)

add further metadata columnsin a spreadsheed application

collect_template(files, include = "*Example_?.csv.xz", outfile = "template.csv")

note the FORMAT:columns are tab separated, fields protected by quotation marks


collect_template()

data frame




save tab separated and use quotation marks as field protector

load file into R environment usingto_metadata()


collect_template()

data frame




save tab separated and use quotation marks as field protector

metadata.example <- to_metadata("template.csv", sep = ",")

metadata.example <- to_metadata("template.csv")metadata.example <- to_metadata("template.csv", strip.white = FALSE)

load file into R environment using


collect_template()

data frame




further added metadata columns

Note: mock metadata for demonstration purpose


example.opm.metadata <- include_metadata(example.opm, md = metadata)

OPMS object with kinetic raw data

data frame with metadatametadata

example.opm

include_metadata()


xy_plot(example.opm)draw kinetic data


xy_plot(example.opm)


xy_plot(example.opm, col = c("blue", "red", "green"), lwd = 2)


xy_plot(example.opm, col = c("blue", "red", "green"), lwd = 2, legend.fmt = list(space = "right"))


xy_plot(example.opm, col = c("blue", "red", "green"), lwd = 2, legend.fmt = list(space = "right"), include = c("Species", "strain"))


xy_plot(example.opm, col = c("blue", "red", "green"), lwd = 2, legend.fmt = list(space = "right"), include = c("Species", "strain"))

Modify panel strip, strip text, and legend by using arguments from lattice


xy_plot(example.opm[plates, time points, wells])

It is possible to plot (1) specific plates, (2) time points, or (3) wellsby indexing OPMS objectsusing square brackets.

xy_plot(example.opm[ , , ])

What about drawing only parts?


xy_plot(example.opm[ , , ])


xy_plot(example.opm[ 3, , ])


xy_plot(example.opm[ 3, 1:100 , ])


xy_plot(example.opm[ 3, 1:100 , c("A01", "A02", "E05", "G08", "H10")])


Heatmaps compare plates on the basis of aggregated curve parameters


The generation of heatmaps includes two steps:

(1) Extract the curve parameter values usingextract()

(2) Create the heatmap usingheat_map()


First step:

AUC <- extract(example.opm, dataframe = TRUE, as.labels = list("country", "Species", "strain", "town"),subset = "AUC")

metadata of interest parameter and values from aggregating the curve parameters


Second step:

heat_map(AUC, as.labels = c("Species", "town"), as.groups = "town", cexRow = 1.2, use.fun = "gplots",main = "nice heatmap",col = topo.colors(120))


heat_map(AUC, as.labels = c("Species", "town"),as.groups = "town",cexRow = 1.2, use.fun = "gplots",main = "nice heatmap",col = topo.colors(120))


Confidence interval plot

Do curves differ significantly in aggregated curve parameters?

We make use of the 95% confidence intervals calculated from 100 bootstrap replicates.



In which aggregated curve parameters

do these curves differ significantly ?


xy_plot(example.opm[ , ,"D10"], include = list("Species","town"), neg.ctrl = FALSE)


ci_plot(example.opm[ , , c("D10")], as.labels = list("Species","town"), subset = "A")


ci_plot(example.opm[ , , c("D10")], as.labels = list("Species","town"), subset = "AUC")


ci_plot(example.opm[ , , c("D10")], as.labels = list("Species","town"), subset = "lambda")



Do these curves differ in their lag phase?

Try yourself


radial_plot(example.opm[, , 5:17], sep = " ", as.labels = c("Species", "town"), draw.legend = FALSE, subset = "AUC")


xy_plot(example.opm[plates, time points, wells])

data(vaas_et_al)

-114 GenIII plates (run 96 hours) - numerous replicates of - each two strains of Escherichia coli and Pseudomonas aeruginosa, - including aggregated bootstrapped curve parameters and metadata


data(vaas_et_al)

An introduction to opm opm – an R package to analyse ... · opm – an R package to analyse...

Documents

Transcript of An introduction to opm opm – an R package to analyse ... · opm – an R package to analyse...