effectR: An R package to call cytoplasmic effectors

What is effectR

The effectR package is an R package designed to call oomycete RxLR and CRN effectors by searching for the motifs of interest using regular expression searches and hidden markov models (HMM).

What’s new in version 1.0.1?

Overview

The effectR packages searches for the motifs of interest (RxLR-EER motif for RxLR effectors and LFLAK motif for CRN effectors) using a regular expression search (REGEX). These motifs used by the REGEX effectR search have been reported in the literature (Haas et al., 2009, Stam et al., 2013).

The effectR package aligns the REGEX search results using MAFFT, and builds a HMM profile based on the multiple sequence alignment result using the hmmbuild program from HMMER. The HMM profile is used to search across ORF of the genome of interest using the hmmsearch binary from HMMER. The search step will retain sequences with significant hits to the profile of interest. effectR also combines the redundant sequences found in both REGEX and HMM searches into a single dataset that can be easily exported. In addition, effectR reads and returns the HMM profile to the user and allows for the creation of a MOTIF logo-like plot using ggplot2.

Requirements

Installation

From CRAN

The latest version of effectR can be installed from CRAN. To install, make sure R is at least version 3.4.0. In the R console type

install.packages("effectR")

From GitHub

To install effectR via GitHub, make sure that the devtools package is installed (use install.packages("devtools")). After installing devtools, in the R console type:

devtools::install_github(repo = "grunwaldlab/effectR", build_vignettes = TRUE)
library("effectR")

External software (REQUIRED)

The effectR package uses MAFFT and HMMER3 to perform the hidden markov model seach across the results from the REGEX step. These two packages should be installed before running any of the effectR functions.

Downloading and installing MAFFT

MAFFT is a multiple sequence alignment program that uses Fourier-transform algorithms to align multiple sequences. We recommend downloading and installing MAFFT by following the instructions and steps in the MAFFT installation web site.

Linux/OS X Users

Make sure that you remember the directory in which MAFFT is installed, of if the installation is sucessful, make sure to obtain the path via bash/tsh/console:

which mafft
/usr/local/bin/mafft

For more information about MAFFT go to the MAFFT website: http://mafft.cbrc.jp/

Windows Users

MAFFT comes in two main distributions for windows:

Please, download and install the all-in-one version. We recommend that you download and save MAFFT in your Desktop, as it will make yyour path easily accesible.

Downloading and installing HMMER

HMMER is used for searching sequence databases for sequence homologs. It uses hidden Markov models (profile HMMs) to search for sequences with hits to similar patterns than the profile. We use three main HMMER tools:

The effectR package requires all of these tools. A correct HMMER installation will install all three programs.

Linux/OS X users

We recommend downloading and installing HMMER by following the instructions and steps in the HMMER installation web site. Make sure that you remember the directory in which HMMER is installed, of if the installation is sucessful, make sure to obtain the path via bash/tsh/console:

which hmmbuild
which hmmpress
which hmmsearch
/usr/local/bin/hmmbuild
/usr/local/bin/hmmpress
/usr/local/bin/hmmsearch

For more information about HMMER go to the HMMER website: http://hmmer.org/

Windows users

To use the effectR package in Windows, the user must download the Windows binaries of HMMER. effectR will not work with any other version of HMMER.
## Data input

The effectR package is designed to work with amino acid sequences in FASTA format representing the six-frame translation of every open reading frame (ORF) of an oomycete genome. Using the six-frame translation of all ORF’s in a genome is recommended in order to obtain as many effectors as possible from a proteome. To obtain the ORF for a genome, we recommend the use of EMBOSS’ getorf.

effectR uses a list of sequences of the class SeqFastadna in order to perform the effector searches. The function read.fasta from the seqinr package reads the FASTA amino acid file into R, creating a list of SeqFastadna objects that represent each of the translated ORF’s from the original FASTA file.

To perform the effector search, effectR searches for the motifs of interest found in RxLR and CRN motifs. We have created the function regex.search to perform the seach of the motif of interest. The function regex.search requires the list of SeqFastadna objects and the gene family of interest.

To perform the HMM search and obtain all possible effector candidates from a proteome, effectR uses the REGEX results as a template to create a HMM profile and perform a search across the proteome of interest. We have created the hmm.search function in order to perfomr this search. The hmm.search function requires a local installation of MAFFT and HMMER in order to perform the searches. The absolute paths of the binaries must be specified in the mafft.path and hmmer.path options of the hmm.search function. In addition, the hmm.function requires the path of the original FASTA file containing the translated ORF’s in the original.seq parameter of the function. hmm.search will use this file as a query in the hmmsearch software from HMMER, and search for all sequences with hits against the HMM profile created with the REGEX results.

A default hmm.search object returns a list of 3 elements:

  1. The REGEX sequences used to build the HMM profile in a SeqFastadna class
  2. The sequences from the original translated ORF files with hits to the HMM profile in a SeqFastadna class
  3. The HMM profile table created by HMMER’s hmmbuild as a data frame

NEW FEATURES: - hmm.search can use a user-defined alignment file (i.e. A multiple sequence alignment performed in MUSCLE, ClustalW, etc.) and omit the alignment step - hmm.search allows the user to save the multiple sequence alignment created by MAFFT within the function

More information on these new features is available in the package help (?hmm.search) or in the effectR vignette

Obtaining non-redundant effectors and motif summaries

The user can extract all of the non-redundant sequences and a summary table with the information about the motifs using the effector.summary function. This function uses the results from either hmm.seach or regex.search functions to generate a table that includes the name of the candidate effector sequence, the number of motifs of interest (RxLR-EER or LFLAK-HVLV) per sequence and its location within the sequence. In addition, when the effector.summary function is used in an object that contains the results of hmm.search, the user will obtain a list of the non-reduntant sequences. If the user provides the results from regex.search, the function will return the motif summary table.

The motif table has a column called MOTIF. This column summarizes the candidate ORF into one of 4 categories:

Non-redundant sequences

To export the non-redundant effector candidates that resulted from the hmm.search or regex.search functions, we use the write.fasta function of the seqinr package. We recomend the users to read the documentation of the seqinr package Since the objects that result from the hmm.search or regex.search function are of the SeqFastadna class, we can use any of the function of the seqinr package that use this class as well.

Visualizing the HMM profile using a sequence logo-like plot

To determine if the HMM profile includes the motifs of interest, we have created the function hmm.logo. The function hmm.logo reads the HMM profile (obtained from the hmm.search step) and uses ggplot2 to create a bar-plot. The bar-plot will illustrate the bits (aminoacid score) of each amino acid used to construct the HMM profile according to its consensus position in the HMM profile. To learn more about sequence logo plots visit this wikipedia article.

Adding custom motifs to effectR

The effectR package has the capability to use custom regular expressions to predict other families of genes of interest other than RxLR/CRN effector proteins. This example uses the PAAR motif (PAAR) identified in proteins associated with the terminal spike in T6SS of some bacterial species:

# Loading the effectR package
library(“effectR”)

# Using the read.fasta function of the seqinr package to import the V. cholerae FASTA proteome file
fasta.file <- " V_cholerae_ATCC_39315.AA.fasta.gz”
ORF <- seqinr::read.fasta(fasta.file)

# Step 1 prediction: Since the PAAR motif can occur anywhere in the sequence of the protein, the REGEX will be very simple and will only contain the PAAR motif.
REGEX <- regex.search(ORF, motif = "custom", reg.pat = "PAAR")

Step 1 resulted in a total of 19 predicted proteins with the PAAR motif. We can expand the number of candidate PAAR proteins using the HMM step:

# Expanding the search of RxLR effectors using HMM searches (step 2). All candidate effectors predicted by step 2 will be saved in the candidate.rxlr object

candidate.paar <- hmm.search(original.seq = fasta.file, regex.seq = REGEX)

Step 2 resulted in one additional candidate protein with a plausible PAAR motif. We can summarize all the information from effectR using the effector.summary() function. It will return a table with the candidate proteins, the number of PAAR motifs within each protein, and the position of said motif:

# Summarizing the results of effectR.

effector.summary(candidate.paar)

                                  Sequence ID Motif number Motif position        MOTIF
tr|Q9KN60|Q9KN60_VIBCH tr|Q9KN60|Q9KN60_VIBCH            2          35,71 Custom motif
sp|Q9KR02|RUVB_VIBCH     sp|Q9KR02|RUVB_VIBCH            1            145 Custom motif
sp|Q9KPV0|GLND_VIBCH     sp|Q9KPV0|GLND_VIBCH            1            398 Custom motif
sp|Q9KSQ2|HUTG_VIBCH     sp|Q9KSQ2|HUTG_VIBCH            1            269 Custom motif
sp|Q9KPU5|NUSB_VIBCH     sp|Q9KPU5|NUSB_VIBCH            1              7 Custom motif
tr|Q9KS85|Q9KS85_VIBCH tr|Q9KS85|Q9KS85_VIBCH            1            171 Custom motif
tr|Q9KUC8|Q9KUC8_VIBCH tr|Q9KUC8|Q9KUC8_VIBCH            1            232 Custom motif
tr|Q9KND6|Q9KND6_VIBCH tr|Q9KND6|Q9KND6_VIBCH            1            153 Custom motif
tr|Q9KN94|Q9KN94_VIBCH tr|Q9KN94|Q9KN94_VIBCH            1            236 Custom motif
tr|Q9KMP0|Q9KMP0_VIBCH tr|Q9KMP0|Q9KMP0_VIBCH            1             35 Custom motif
tr|Q9KPU1|Q9KPU1_VIBCH tr|Q9KPU1|Q9KPU1_VIBCH            1            179 Custom motif
tr|Q9KSK0|Q9KSK0_VIBCH tr|Q9KSK0|Q9KSK0_VIBCH            1            501 Custom motif
tr|Q9KLU5|Q9KLU5_VIBCH tr|Q9KLU5|Q9KLU5_VIBCH            1            481 Custom motif
tr|Q9KKR8|Q9KKR8_VIBCH tr|Q9KKR8|Q9KKR8_VIBCH            1            343 Custom motif
tr|Q9KPP4|Q9KPP4_VIBCH tr|Q9KPP4|Q9KPP4_VIBCH            1            811 Custom motif
tr|Q9KUF6|Q9KUF6_VIBCH tr|Q9KUF6|Q9KUF6_VIBCH            1            290 Custom motif
tr|Q9KVN5|Q9KVN5_VIBCH tr|Q9KVN5|Q9KVN5_VIBCH            1             74 Custom motif
tr|Q9KQJ5|Q9KQJ5_VIBCH tr|Q9KQJ5|Q9KQJ5_VIBCH            1            189 Custom motif
tr|Q9KNT2|Q9KNT2_VIBCH tr|Q9KNT2|Q9KNT2_VIBCH            1            146 Custom motif
tr|Q9KUM6|Q9KUM6_VIBCH tr|Q9KUM6|Q9KUM6_VIBCH            0           <NA>    No MOTIFS

The results illustrate that 19 out of the 20 candidate proteins have a predicted PAAR domain within its sequence, and only protein tr|Q9KN60|Q9KN60_VIBCH has more than 1 PAAR motif.

Any user can add the motif of interest into the effectR package by adding a simple line of code within the regex.search function. We will illustrate this feature by adding the PAAR motif search as part of the regex.search function:

# The regex.search function
regex.search <- function(sequence, motif = "RxLR", reg.pat = NULL){
 if (unique(unlist(lapply(sequence, class))) != "SeqFastadna") {
    stop("The object is not a list of sequences read by seqinr.")
  }
  seq <- lapply(sequence, function (x) paste(unlist(x),collapse = ""))
    regex <- list()
    if (motif %in% c("RxLR","CRN",”PAAR”) & !is.null(reg.pat)){
      message(paste0("Custom REGEX patterns are not supported with the 'CRN' or 'RxLR' motif options.\n The package will use the default REGEX patterns used to search for ", motif, " motifs."))
      Sys.sleep(2)
    }
    for (i in 1:length(seq)){
      if (motif == "RxLR"){
        reg.pat <- "^\\w{10,40}\\w{1,96}R\\wLR\\w{1,40}eer"
      } else if (motif == "CRN"){
        reg.pat <- "^\\w{1,90}LFLAK\\w+"
      } else if (motif == "PAAR"){
        reg.pat <- “PAAR”
    } else if (motif == "custom"){
        if (is.null(reg.pat)){
          stop("No custom REGEX pattern found.\n The 'custom' option requires a mandatory REGEX pattern")
        } else {
        reg.pat <- reg.pat
        }
      }
      regex[[i]] <- unlist(gregexpr(seq[[i]], pattern = reg.pat, perl = T ,ignore.case = T))
    #percentage <- percentage + 1/length(seq)*100
    }
    regex <- as.data.frame(do.call(rbind, regex))
    regex$seq <- names(seq)
    regex <- regex[!regex$V1 < 0, ]
    regex <- sequence[seqinr::getName(sequence) %in% regex$seq]
    if (length(regex) == 0){
      stop(paste0("No ",motif, " sequences found."))
    }
    return(regex)
}

After including the new PAAR motif, the user can specify the PAAR motif as an option of the motif parameter:

# Loading the effectR package in R
library(“effectR”)

# Using the read.fasta function of the seqinr package to import the V. cholerae FASTA proteome file

fasta.file <- " V_cholerae_ATCC_39315.AA.fasta.gz”
ORF <- seqinr::read.fasta(fasta.file)

# Step 1 prediction: Predict proteins with the PAAR motif

REGEX <- regex.search(ORF, motif = "PAAR)

This customization will allow users to add any motif of interest to the effectR package by forking the github repository, adding the additional line into the regex.search function, adding the respective reference of the motif into the @references section of the R documentation within the regex.search function, and submitting a pull request to the package maintainers. The package maintainers will update the package, test the motif, and, if valid, add the motif to the effector.summary function before updating effectR. The currently available CRAN version of effectR only includes the RxLR and CRN motifs to facilitate the familiarization and engagement of the community with the package, but additional custom REGEX patterns will be added as the package is updated.