In snystrom/memes: motif matching, comparison, and de novo discovery using the MEME Suite

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>",
  fig.path = "man/figures/README-",
  fig.align = "center"
  #out.width = "100%"
)

memes

An R interface to the MEME Suite family of tools, which provides several utilities for performing motif analysis on DNA, RNA, and protein sequences. memes works by detecting a local install of the MEME suite, running the commands, then importing the results directly into R.

Installation

Bioconductor

memes is currently available on the Bioconductor devel branch:

if (!requireNamespace("BiocManager", quietly = TRUE))
    install.packages("BiocManager")

# The following initializes usage of Bioc devel
BiocManager::install(version='devel')

BiocManager::install("memes")

Development Version (Github)

You can install the development version of memes from GitHub with:

if (!requireNamespace("remotes", quietly=TRUE))
  install.packages("remotes")
remotes::install_github("snystrom/memes")

# To temporarily bypass the R version 4.1 requirement, you can pull from the following branch:
remotes::install_github("snystrom/memes", ref = "no-r-4")

Docker Container

# Get development version from dockerhub
docker pull snystrom/memes_docker:devel
# the -v flag is used to mount an analysis directory, 
# it can be excluded for demo purposes
docker run -e PASSWORD=<password> -p 8787:8787 -v <path>/<to>/<project>:/mnt/<project> snystrom/memes_docker:devel

Detecting the MEME Suite

memes relies on a local install of the MEME Suite. For installation instructions for the MEME suite, see the MEME Suite Installation Guide.

memes needs to know the location of the meme/bin/ directory on your local machine. You can tell memes the location of your MEME suite install in 4 ways. memes will always prefer the more specific definition if it is a valid path. Here they are ranked from most- to least-specific:

1. Manually passing the install path to the meme_path argument of all memes functions
2. Setting the path using options(meme_bin = "/path/to/meme/bin/") inside your R script
3. Setting MEME_BIN=/path/to/meme/bin/ in your .Renviron file
4. memes will try the default MEME install location ~/meme/bin/

If memes fails to detect your install at the specified location, it will fall back to the next option.

To verify memes can detect your MEME install, use check_meme_install() which uses the search herirarchy above to find a valid MEME install. It will report whether any tools are missing, and print the path to MEME that it sees. This can be useful for troubleshooting issues with your install.

library(memes)

# Verify that memes detects your meme install
# (returns all green checks if so)
check_meme_install()

# You can manually input a path to meme_path
# If no meme/bin is detected, will return a red X
check_meme_install(meme_path = 'bad/path')

The Core Tools

| Function Name | Use | Sequence Input | Motif Input | Output | |:-------------:|:----------------:|:--------------:|:-----------:|:-------------------------------------------------------| | runStreme() | Motif Discovery (short motifs) | Yes | No | universalmotif_df | | runDreme() | Motif Discovery (short motifs) | Yes | No | universalmotif_df | | runAme() | Motif Enrichment | Yes | Yes | data.frame (optional: sequences column) | | runFimo() | Motif Scanning | Yes | Yes | GRanges of motif positions | | runTomTom() | Motif Comparison | No | Yes | universalmotif_df w/ best_match_motif and tomtom columns* | | runMeme() | Motif Discovery (long motifs) | Yes | No | universalmotif_df |

* Note: if runTomTom() is run using a universalmotif_df the results will be joined with the universalmotif_df results as extra columns. This allows easy comparison of de-novo discovered motifs with their matches.

Sequence Inputs can be any of:

1. Path to a .fasta formatted file
2. Biostrings::XStringSet (can be generated from GRanges using get_sequence() helper function)
3. A named list of Biostrings::XStringSet objects (generated by get_sequence())

Motif Inputs can be any of:

1. A path to a .meme formatted file of motifs to scan against
2. A universalmotif object, or list of universalmotif objects
3. A runDreme() results object (this allows the results of runDreme() to pass directly to runTomTom())
4. A combination of all of the above passed as a list() (e.g. list("path/to/database.meme", "dreme_results" = dreme_res))

Output Types:

runDreme(), runStreme(), runMeme() and runTomTom() return universalmotif_df objects which are data.frames with special columns. The motif column contains a universalmotif object, with 1 entry per row. The remaining columns describe the properties of each returned motif. The following column names are special in that their values are used when running update_motifs() and to_list() to alter the properties of the motifs stored in the motif column. Be careful about changing these values as these changes will propagate to the motif column when calling update_motifs() or to_list().

• name
• altname
• family
• organism
• strand
• nsites
• bkgsites
• pval
• qval
• eval

memes is built around the universalmotif package which provides a framework for manipulating motifs in R. universalmotif_df objects can interconvert between data.frame and universalmotif list format using the to_df() and to_list() functions, respectively. This allows use of memes results with all other Bioconductor motif packages, as universalmotif objects can convert to any other motif type using convert_motifs().

runTomTom() returns a special column: tomtom which is a data.frame of all match data for each input motif. This can be expanded out using tidyr::unnest(tomtom_results, "tomtom"), and renested with nest_tomtom(). The best_match_ prefixed columns returned by runTomTom() indicate values for the motif which was the best match to the input motif.

Quick Examples

Motif Discovery with DREME

suppressPackageStartupMessages(library(magrittr))
suppressPackageStartupMessages(library(GenomicRanges))

# Example transcription factor peaks as GRanges
data("example_peaks", package = "memes")

# Genome object
dm.genome <- BSgenome.Dmelanogaster.UCSC.dm6::BSgenome.Dmelanogaster.UCSC.dm6

The get_sequence function takes a GRanges or GRangesList as input and returns the sequences as a BioStrings::XStringSet, or list of XStringSet objects, respectively. get_sequence will name each fasta entry by the genomic coordinates each sequence is from.

# Generate sequences from 200bp about the center of my peaks of interest
sequences <- example_peaks %>% 
  resize(200, "center") %>% 
  get_sequence(dm.genome)

runDreme() accepts XStringSet or a path to a fasta file as input. You can use other sequences or shuffled input sequences as the control dataset.

# runDreme accepts all arguments that the commandline version of dreme accepts
# here I set e = 50 to detect motifs in the limited example peak list
# In a real analysis, e should typically be < 1
dreme_results <- runDreme(sequences, control = "shuffle", e = 50)

memes is built around the universalmotif package. The results are returned in universalmotif_df format, which is an R data.frame that can seamlessly interconvert between data.frame and universalmotif format using to_list() to convert to universalmotif list format, and to_df() to convert back to data.frame format. Using to_list() allows using memes results with all universalmotif functions like so:

library(universalmotif)

dreme_results %>% 
  to_list() %>% 
  view_motifs()

Matching motifs using TOMTOM

Discovered motifs can be matched to known TF motifs using runTomTom(), which can accept as input a path to a .meme formatted file, a universalmotif list, or the results of runDreme().

TomTom uses a database of known motifs which can be passed to the database parameter as a path to a .meme format file, or a universalmotif object.

Optionally, you can set the environment variable MEME_DB in .Renviron to a file on disk, or the meme_db value in options to a valid .meme format file and memes will use that file as the database. memes will always prefer user input to the function call over a global variable setting.

options(meme_db = system.file("extdata/flyFactorSurvey_cleaned.meme", package = "memes"))
m <- create_motif("CMATTACN", altname = "testMotif")
tomtom_results <- runTomTom(m)

tomtom_results

Using runDreme results as TOMTOM input

runTomTom() will add its results as columns to a runDreme() results data.frame.

full_results <- dreme_results %>% 
  runTomTom()

Motif Enrichment using AME

AME is used to test for enrichment of known motifs in target sequences. runAme() will use the MEME_DB entry in .Renviron or options(meme_db = "path/to/database.meme") as the motif database. Alternately, it will accept all valid inputs similar to runTomTom().

# here I set the evalue_report_threshold = 30 to detect motifs in the limited example sequences
# In a real analysis, evalue_report_threshold should be carefully selected
ame_results <- runAme(sequences, control = "shuffle", evalue_report_threshold = 30)
ame_results

Visualizing Results

view_tomtom_hits allows comparing the input motifs to the top hits from TomTom. Manual inspection of these matches is important, as sometimes the top match is not always the correct assignment. Altering top_n allows you to show additional matches in descending order of their rank.

full_results %>% 
  view_tomtom_hits(top_n = 1)

It can be useful to view the results from runAme() as a heatmap. plot_ame_heatmap() can create complex visualizations for analysis of enrichment between different region types (see vignettes for details). Here is a simple example heatmap.

ame_results %>% 
  plot_ame_heatmap()

Scanning for motif occurances using FIMO

The FIMO tool is used to identify matches to known motifs. runFimo will return these hits as a GRanges object containing the genomic coordinates of the motif match.

# Query MotifDb for a motif
e93_motif <- MotifDb::query(MotifDb::MotifDb, "Eip93F") %>% 
  universalmotif::convert_motifs()

# Scan for the E93 motif within given sequences
fimo_results <- runFimo(sequences, e93_motif, thresh = 1e-3)

# Visualize the sequences matching the E93 motif
plot_sequence_heatmap(fimo_results$matched_sequence)  

Importing Data from previous runs

memes also supports importing results generated using the MEME suite outside of R (for example, running jobs on meme-suite.org, or running on the commandline). This enables use of preexisting MEME suite results with downstream memes functions.

| MEME Tool | Function Name | File Type | |:---------:|:-------------------:|:----------------:| | Streme | importStremeXML() | streme.xml | | Dreme | importDremeXML() | dreme.xml | | TomTom | importTomTomXML() | tomtom.xml | | AME | importAme() | ame.tsv* | | FIMO | importFimo() | fimo.tsv | | Meme | importMeme() | meme.txt |

* importAME() can also use the "sequences.tsv" output when AME used method = "fisher", this is optional.

FAQs

How do I use memes/MEME on Windows?

The MEME Suite does not currently support Windows, although it can be installed under Cygwin or the Windows Linux Subsytem (WSL). Please note that if MEME is installed on Cygwin or WSL, you must also run R inside Cygwin or WSL to use memes.

An alternative solution is to use Docker to run a virtual environment with the MEME Suite installed. We provide a memes docker container
that ships with the MEME Suite, R studio, and all memes dependencies pre-installed.

Citation

memes is a wrapper for a select few tools from the MEME Suite, which were developed by another group. In addition to citing memes, please cite the MEME Suite tools corresponding to the tools you use.

If you use runDreme() in your analysis, please cite:

Timothy L. Bailey, "DREME: Motif discovery in transcription factor ChIP-seq data", Bioinformatics, 27(12):1653-1659, 2011. full text

If you use runTomTom() in your analysis, please cite:

Shobhit Gupta, JA Stamatoyannopolous, Timothy Bailey and William Stafford Noble, "Quantifying similarity between motifs", Genome Biology, 8(2):R24, 2007. full text

If you use runAme() in your analysis, please cite:

Robert McLeay and Timothy L. Bailey, "Motif Enrichment Analysis: A unified framework and method evaluation", BMC Bioinformatics, 11:165, 2010, doi:10.1186/1471-2105-11-165. full text

If you use runFimo() in your analysis, please cite:

Charles E. Grant, Timothy L. Bailey, and William Stafford Noble, "FIMO: Scanning for occurrences of a given motif", Bioinformatics, 27(7):1017-1018, 2011. full text

Licensing Restrictions

The MEME Suite is free for non-profit use, but for-profit users should purchase a license. See the MEME Suite Copyright Page for details.

snystrom/memes documentation built on Oct. 12, 2024, 2:42 a.m.