In thomasp85/MSsary: LC-MS/MS analysis in R

---

Introduction

MSsary was created with an aim to please both users and developers. If you are reading this I assume that you fall comfortably into the latter category. The strategy with MSsary has been to create a framework so compelling that people will want to implement the next best algorithm in it, so while the package ships with a range of algorithms to get the user started, the hope is that with time algorithms will be ported from other open source project as well as developed from the ground up. To support this MSsary has defined a very simple system for plugging in new algorithms in the framework that frees developers from thinking about interfacing with the underlying sqlite architecture and focus on working with the data of their interest. To underline the commitment to extensibility in MSsary all the internal algorithms are also implemented as extensions. At present the following parts of MSsary are open to extensions:

• Preprocessing of scans e.g. centroiding, filtering, recalibration
• Detection of chromatographic peaks
• Linking of peaks e.g. isotope and adduct detection, label detection
• Grouping of peaks across samples

Furthermore there is the possibility to augment the metadata available for the different main data types, such as assigning a quality score to all scans, or calculate parameters for a gaussian fit to a peak.

In the following the general mechanism for extending MSsary with your own package will be described and all specific extensions from the above list will be adressed in detail.

The extension framework

In very general terms all types of algorithms are handled by their own reference class object that takes care of checking input, output, method conformity etc. These objects are also responsible for registering new methods. Usually it is advisable to use a package as a container for new algorithms, as it makes it possible for MSsary to keep track of version history of the algorithms etc. Assuming that your new algorithm is distributed in a package, the registering should be done in onLoad(). Below is an example of how massifquant is registered:

.onLoad <- function(libname, pkgname) {
    peakMethods$registerMethod(
        name='massifquant', 
        fun=massifquant, 
        req=list(mode='centroid'))
}

peakMethods is the reference class object that takes care of handling all peak detection algorithms. The registerMethod method is used for registering new peak detection algorithms and takes three parameters: name is the name of the algorithm to be used internally, fun is the actual function from your package and req is a list of requirements for the input that should be checked before running the algorithm. Each type of algorithm have different types of requirements that can be set as well as different specifications for the parameters of fun. These will be discussed below under their respective sections.

thomasp85/MSsary documentation built on May 31, 2019, 11:11 a.m.