robertoalvarezm / 18s-pipeline-dada2-r

Pipeline for running DADA2 in R specific for microbial eukaryotic work.

18s-pipeline-dada2-R

Pipeline for running DADA2 in R specific for microbial eukaryotic work.

(1) Perform initial QC & removal primers

For high-throughput sequence quality control and primer removal, I use snakemake.

# build conda environment to run snakemake
conda env create --name snake-qc --file envs/snake.yaml

# Enter this conda environment
conda activate snake-qc

Set up working directory

1. Modify your config.yaml file to tell snakemake where to look for raw sequences, the format of those sequences, and where you want the output trimmed reads. Use a text editor like nano to modify this file.
2. Raw fastq sequences should be

(2) DADA2 pipeline in R

library(dada2)

## pending ##

---

Assign taxonomy to reference sequences

Use assignTaxonomy() from the dada2 package to assign taxonomy. Below code can also be used to assign taxonomy to other sequences.

(1) Bring sequences for assignment into R

Extract reference sequences from qiime2 run:

# Enter qiime2 environment
qiime tools export --input-path ref-sequences.qza  
	--output-path ref-seqs/

Output will be a new directory called ref-seqs/, inside is a .fna file with all reference sequences. Fasta headers are Feature.IDs from output ASV or OTU table.

Now import those reads into R to assign taxonomy from dada2.

## R v3.6.1

# Read in reference sequence file, create dataframe with Feature.ID and reference sequence as columns

library(Biostrings)
fna_in <- readDNAStringSet("xxxx.fasta") #Import reference sequences
Feature.ID <- names(fna_in)
SEQUENCE <- paste(fna_in)
fna_df <- data.frame(Feature.ID, SEQUENCE)
# Dataframe with 1 column of sequence header and a 2nd column with sequence.

(2) Assign taxonomy

library(dada2)

seqs <- as.character(fna_df$SEQUENCE) #extract sequences

# Assign taxonomy. Note that for the PR2 database in dada2, you need to use the taxLevels argument as is here (do not use the default)
taxa_pr2 <- assignTaxonomy(seqs, "/vortexfs1/omics/huber/shu/db/pr2-db/pr2_version_4.12.0_18S_dada2.fasta.gz", 
	taxLevels = c("Kingdom","Supergroup","Division","Class","Order","Family","Genus","Species"), 
	multithread = TRUE, minBoot = 0, outputBootstraps = TRUE)

# See "Compile-taxonomy-assignments.ipynb" to compile with exisiting count table

(3) Compile with original fasta file or count table

See R notebook Compile-taxonomy-assignments.ipynb to compile taxonomy assignments with previous count table. This specific example compared output from qiime2 taxonomy and dada2 approach.

library(Biostrings); library(tidyverse)
## ASVs
ref_asv <- readDNAStringSet("/vortexfs1/omics/huber/shu/slo-pier-weekly/qiime2/asv/slo-pier-ref-seqs-asv.fna")
Feature.ID <- names(ref_asv)
ReferenceSequence <- paste(ref_asv)
fna_df <- data.frame(Feature.ID, ReferenceSequence)

load("Pier-assigned-refseqs.RData", verbose = T)
asv_table <- read.delim("/vortexfs1/omics/huber/shu/slo-pier-weekly/qiime2/asv/CountTable-wtax-2020-04-22.txt")

# Comile ASV results with reference sequence
asv_wtax <- data.frame(taxa_pr2) %>% 
    rownames_to_column(var = "ReferenceSequence") %>% 
    right_join(fna_df) %>% 
    unite(Taxon_dada2_boot0, starts_with("tax."), sep = ";") %>% 
    unite(Confidence_dada2, starts_with("boot."), sep = ";") %>% 
    left_join(asv_table) %>% 
    select(Feature.ID, Taxon_qiime2 = Taxon, Conf_qiime2 = Confidence, 
            Taxon_dada2_boot0, Conf_dada2_boot0 = Confidence_dada2, everything()) %>% 
    data.frame
# head(asv_wtax)

(4) Explore & reassign taxonomy from PR2

Using the list of taxa and bootstrap values for all taxonomic assignment levels, we need to determine what the most appropriate threshold is.

In R (v3.6.1)

library(tidyverse)

# Import R object that has FeatureID information, reference sequence, and PR2 assignment information from above
head(asv_wtax)


# Create new column of updated taxonomic list, based on a minBoot threshold of 70
asv_updated_tax <- asv_wtax %>% 
  type.convert(as.is = TRUE) %>%
  separate(Confidence_dada2, c("Kingdom_boot","Supergroup_boot","Division_boot","Class_boot","Order_boot","Family_boot","Genus_boot","Species_boot"), sep = ";", convert = TRUE) %>%
  separate(Taxon_dada2_boot0, c("Kingdom_lev","Supergroup_lev","Division_lev","Class_lev","Order_lev","Family_lev","Genus_lev","Species_lev"), sep = ";", convert = TRUE) %>%
  mutate(Taxon_updated = 
     case_when(
      Species_boot >= 70 ~
         select(., ends_with("_lev")) %>% 
            reduce(str_c, sep=";"),
      Species_boot < 70 & Genus_boot >= 70 ~ 
          select(., Kingdom_lev:Genus_lev) %>%
            reduce(str_c, sep=";"),
      Genus_boot < 70 & Family_boot >=70 ~
          select(., Kingdom_lev:Family_lev) %>%
            reduce(str_c, sep=";"),
      Family_boot < 70 & Order_boot >=70 ~
          select(., Kingdom_lev:Order_lev) %>%
            reduce(str_c, sep=";"),
      Order_boot < 70 & Class_boot >=70 ~
          select(., Kingdom_lev:Class_lev) %>%
            reduce(str_c, sep=";"),
      Class_boot < 70 & Division_boot >=70 ~
          select(., Kingdom_lev:Division_lev) %>%
            reduce(str_c, sep=";"),
      Division_boot < 70 & Supergroup_boot >=70 ~
          select(., Kingdom_lev:Supergroup_lev) %>%
            reduce(str_c, sep=";"),
      TRUE ~ Kingdom_lev)) %>% 
  data.frame

Output generates a new column with an updated taxonomic assignment that is concatenated (;) by the appropriate minimum bootstrap value. Option to set different levels for various taxonomic levels.

Last updated - 19-08-2020 - SKH