Introduction to Using RSeqAn

August Guang

2019-02-15

Introduction

The reason RSeqAn was created was to allow for easy integration of the SeqAn biological sequence analysis C++ library into R packages. While R is an excellent language for many other applications, it is just not fast enough for reading and writing files on the scale of next generation sequencing output. This is where a well-developed and mature library like SeqAn comes in.

This vignette only goes through the first example in the A First Example section as found in the Getting Started section of the SeqAn docs. We have modified the function slightly to make it work here in R, and will go through how and why we did so. The purpose of using this example is to help the user get an idea of how to go between SeqAn and R. To take full advantage of SeqAn though the user will need to read through SeqAn’s documentation.

Besides that, the user is expected to have some experience with both C++ and Rcpp, although it not need be extensive. After all, that is what RSeqAn is for.

Template functions and template classes

The simple example in pattern_search does, as you might expect, a pattern search of a short query sequence (pattern) in a long subject sequence. It returns an integer score value for each position of the database sequence (text) as the sum of matching characters between the pattern string and the subject substring of the database sequence.

// [[Rcpp::depends(RSeqAn)]]

#include <iostream>
#include <seqan/file.h>
#include <seqan/sequence.h>
#include <Rcpp.h>

using namespace Rcpp;
using namespace seqan;

// [[Rcpp::export]]
IntegerVector pattern_search(std::string t, std::string p) {
    
    seqan::String<char> text = t;
    seqan::String<char> pattern = p;
    
    seqan::String<int> score;
    resize(score, length(text) - length(pattern) + 1);
    
    // Computation of the similarities
    // Iteration over the text (outer loop)
    for (unsigned i = 0; i < length(text) - length(pattern) + 1; ++i)
    {
        int localScore = 0;
        // Iteration over the pattern for character comparison
        for (unsigned j = 0; j < length(pattern); ++j)
        {
            if (text[i + j] == pattern[j])
                ++localScore;
        }
        score[i] = localScore;
    }
    
    // Returning the result
    IntegerVector s(length(score));
    for (unsigned i = 0; i < length(score); ++i)
        s[i] = score[i];
    
    return s;
}

The results are shown below. We see that the maximum score possible is 8, as there are 8 characters in the pattern tutorial, and it achieves that maximum score when we match together tutorial in the text string and tutorial in the pattern string. As well, the first position has a score of 1, because the i in the pattern string tutorial matches the 1 i in is; the subject substring here is This is.

pattern_search("This is an awesome tutorial to get to know SeqAn!", "tutorial")
#>  [1] 1 0 1 0 0 1 0 0 0 0 0 0 1 0 0 0 0 1 0 8 0 1 0 0 0 0 2 0 1 0 0 1 0 3 0
#> [36] 1 0 1 0 0 0 0

A more detailed look at the program

As we can see, writing a C++ function that utilizes SeqAn inside R is quite easy with Rcpp. We included <seqan/file.h> as well as <seqan/sequence.h> as those are the modules that provide the SeqAn String class. This is one of the most fundamental classes in SeqAn.

However, the function we wrote does look slightly different from the one in the A First Example section. First, instead of the function int main(), we have instead written the function IntegerVector pattern_search(std::string t, std::string p). (Note: we already declared the namespace std, but it was left here in the function for clarity) Next, instead of printing the score to stdout, we are returning it as an IntegerVector.

The reason we did this is that in order for any function using SeqAn to be useful in R, we probably want it to return something and to take in input. This means that the input and output object types need to be translatable between R and C++. SeqAn uses its own template functions and template classes, and the String class is one of the most fundamental classes in SeqAn. This makes sense since SeqAn is all about analyzing sequences. However, the String class has no direct translation to R. If you try to input String<char> text or return String<int> score you will end up with loads of errors from the compiler. So, how do we deal with this?

One way to do this is by writing conversion functions such that R and C++ both understand what the data type you are using (such as String) means. Rcpp provides a nice way to do this through Rcpp::as<T>(obj) to convert from R to C++ and Rcpp::wrap(obj) to convert from C++ to R. More of this is covered in the Rcpp vignette Extending Rcpp. Once these functions are written, this is nice for the user as they can just go ahead and Rcpp::wrap and Rcpp::as<T> as they need. This has not been implemented in RSeqAn yet though, although it is a future goal. Thus for now the user will have to pay attention to how to convert between classes in SeqAn and objects in R for each function that is written.

Rcpp has its own data types for going between R and C++, and so that is the IntegerVector we declare here. Since score is essentially a vector of class String with type int, instead of iterating through score and printing to stdout, we create an IntegerVector s with the same length as score and iterate through score copying its values to s in order to be able to return the values in score. Similarly, we make use of the fact that Rcpp already autoconverts character strings in R to character strings in C++ and that character strings in C++ can be converted to String<char> in SeqAn to write pattern_search such that we can run it from R.