README.md

R-CMD-check codecov

mbtools is a collection of helpers that we use to analyze microbiome data. It makes it easier to run some common analyses and is pretty opinionated towards our own experiences.

General philosophy

mbtools sees analyses as a workflow consisting of several analysis steps and final outputs based on those. This is pretty similar to Qiime 2 which pioneered that abstraction.

Data types

mbtools mostly works on the following data types:

  1. artifacts - compound data objects returned from an analysis step
  2. phyloseq object - a phyloseq object containing sequence variant abundances, taxonomy assignments and sample metadata
  3. data tables - general data frame-like objects in tidy format

Workflow steps

For mbtools a workflow step is based on input data and a configuration, thus having the function signature step(object, config). Most steps can be chained with the pipe operator to yield workflows. For instance, the following is a possible workflow with mbtools:

library(mbtools)

config <- list(
    demultiplex = config_demultiplex(barcodes = c("ACGTA", "AGCTT")),
    preprocess = config_preprocess(truncLen = 200),
    denoise = config_denoise()
)

output <- find_read_files("raw") %>%
          demultiplex(config$demultiplex) %>%
          quality_control() %>%
          preprocess(config$preprocess) %>%
          denoise(config$denoise)

This clearly logs the used workflow and the configuration. The configuration can also be saved and read in many formats, for instance yaml.

config.yaml:

preprocess:
  threads: yes
  out_dir: preprocessed
  trimLeft: 10.0
  truncLen: 200.0
  maxEE: 2.0
denoise:
  threads: yes
  nbases: 2.5e+08
  pool: no
  bootstrap_confidence: 0.5
  taxa_db: https://zenodo.org/record/1172783/files/silva_nr_v132_train_set.fa.gz?download=1
  species_db: https://zenodo.org/record/1172783/files/silva_species_assignment_v132.fa.gz?download=1
  hash: yes

This can now be reused by someone else:

config <- read_yaml("config.yml")

output <- find_read_files("raw") %>%
          quality_control() %>%
          preprocess(config$preprocess) %>%
          denoise(config$denoise)

Other functions

All other functions are usually functions that are meant to be inside more complex code or functions that produce plots and endpoints of an analysis. Most of them act on phyloseq objects and some on tidy data tables.



Gibbons-Lab/mbtools documentation built on Jan. 28, 2024, 11:08 a.m.