OTU picking and Denoising

Usually, amplicon sequences are clustered into Operational Taxonomic Units (OTUs) using a similarity threshold of 97%, which represents the common working definition of bacterial species.

Another approach consists to define the so-called Sequence Variants (SVs, a.k.a Amplicon Sequence Variants - ASVs, Exact Sequence Variants ESVs, zero-radius OTUs - ZOTU, unique sequence variants or sub-OTUs). This approach avoids clustering sequences at a predefined similarity threshold and usually includes a denoising algorithm in order to identify SVs (see UNOISE, DADA2, Deblur, oligotyping and swarm).

The otu command assigns similar sequences (marker genes such as 16S rRNA and the fungal ITS region) to operational taxonomic units or sequence variants (OTUs or SVs).

Methods

De novo greedy

In denovo greedy clustering (parameter --method denovo_greedy), sequences are clustered without relying on an external reference database, using an approach similar to the UPARSE pipeline (https://doi.org/10.1038/nmeth.2604) and tested in https://doi.org/10.7287/peerj.preprints.1466v1. This protocol includes in a single command dereplication, clustering and chimera filtering:

  1. Dereplication. Predict sequence abundances of each sequence by dereplication, order by abundance and discard sequences with abundance value smaller than MINSIZE (option -s/--minsize, default value 2);
  2. Greedy clustering. Distance (DGC) and abundance-based (AGC) strategies are supported (option --greedy, see https://doi.org/10.1186/s40168-015-0081-x and https://doi.org/10.7287/peerj.preprints.1466v1 ). Therefore, the candidate representative sequences are obtained;
  3. Chimera filtering (optional). Remove chimeric sequences from the representatives performing a de novo chimera detection (option --rmchim);
  4. Map sequences. Map sequences to the representatives.

Example:

micca otu -m denovo_greedy -i filtered.fasta -o denovo_greedy_otus -d 0.97 -c -t 4

De novo UNOISE

Denoise amplicon sequences using the UNOISE3 protocol. The method is designed for Illumina (paired or unpaired) reads. This protocol includes in a single command dereplication, denoising and chimera filtering:

  1. Dereplication; Predict sequence abundances of each sequence by dereplication, order by abundance and discard sequences with abundance value smaller than MINSIZE (option -s/--minsize, default value 8);
  2. Denoising;
  3. Chimera filtering (optional);
  4. Map sequences. Map sequences to the representatives.

Example:

micca otu -m denovo_unoise -i filtered.fasta -o denovo_unoise_otus -c -t 4

Closed-reference

Sequences are clustered against an external reference database and reads that could not be matched are discarded (method closed_ref). Example:

Download the reference database (Greengenes), clustered at 97% identity:

wget ftp://ftp.fmach.it/metagenomics/micca/dbs/gg_2013_05.tar.gz
tar -zxvf gg_2013_05.tar.gz

Run the closed-reference protocol:

micca otu -m closed_ref -i filtered.fasta -o closed_ref_otus -r 97_otus.fasta -d 0.97 -t 4

Simply perform a sequence ID matching with the reference taxonomy file (see classify):

cd closed_ref_otus
micca classify -m otuid -i otuids.txt -o taxa.txt -x ../97_otu_taxonomy.txt

Open-reference

Open-reference clustering (open_ref): sequences are clustered against an external reference database (as in Closed-reference) and reads that could not be matched are clustered with the De novo greedy protocol. Example:

Download the reference database (Greengenes), clustered at 97% identity:

wget ftp://ftp.fmach.it/metagenomics/micca/dbs/gg_2013_05.tar.gz
tar -zxvf gg_2013_05.tar.gz

Run the open-reference protocol:

micca otu -m open_ref -i filtered.fasta -o open_ref_otus -r 97_otus.fasta -d 0.97 -t 4 -c

Run the VSEARCH-based consensus classifier or the RDP classifier (see classify):

cd open_ref_otus
micca classify -m cons -i otus.fasta -o taxa.txt -r ../97_otus.fasta -x ../97_otu_taxonomy.txt -t 4

De novo swarm

In denovo swarm clustering (doi: 10.7717/peerj.593, doi: 10.7717/peerj.1420, https://github.com/torognes/swarm, parameter --method denovo_swarm), sequences are clustered without relying on an external reference database. From https://github.com/torognes/swarm:

The purpose of swarm is to provide a novel clustering algorithm that handles massive sets of amplicons. Results of traditional clustering algorithms are strongly input-order dependent, and rely on an arbitrary global clustering threshold. swarm results are resilient to input-order changes and rely on a small local linking threshold d, representing the maximum number of differences between two amplicons. swarm forms stable, high-resolution clusters, with a high yield of biological information.

otu includes in a single command dereplication, clustering and de novo chimera filtering:

  1. Dereplication. Predict sequence abundances of each sequence by dereplication, order by abundance and discard sequences with abundance value smaller than MINSIZE (option --minsize default value is 1, i.e. no filtering);
  2. Swarm clustering. Fastidious option is recommended (--swarm-fastidious);
  3. Chimera filtering (optional).

Warning

Removing ambiguous nucleotides (N) (with the option --maxns 0 in filter) is mandatory if you use the de novo swarm clustering method.

Example:

micca filter -i trimmed.fastq -o filtered.fasta -e 0.5 -m 350 -t --maxns 0
micca otu -m denovo_swarm -i filtered.fasta -o otus_denovo_swarm -c --minsize 1 --swarm-fastidious -t 4

Definition of identity

In micca, the pairwise identity (except for ‘de novo swarm’ and ‘denovo unoise’) is defined as the edit distance excluding terminal gaps (same as in USEARCH and BLAST):

\frac{\textrm{\# matching columns}}{\textrm{alignment length} - \textrm{terminal gaps}}

Output files

The otu command returns in a single directory 5 files:

otutable.txt

TAB-delimited file, containing the number of times an OTU is found in each sample (OTU x sample, see Supported file formats):

OTU     Mw_01 Mw_02 Mw_03 ...
DENOVO1 151   178   177   ...
DENOVO2 339   181   142   ...
DENOVO3 533   305   63    ...
...     ...   ...   ...   ...
otus.fasta

FASTA containing the representative sequences (OTUs):

>DENOVO1
GACGAACGCTGGCGGCGTGCCTAACACATGCAAGTCGAACGGGG...
>DENOVO2
GATGAACGCTAGCTACAGGCTTAACACATGCAAGTCGAGGGGCA...
>DENOVO3
AGTGAACGCTGGCGACGTGGTTAAGACATGCAAGTCGAGCGGTA...
...
otuids.txt

TAB-delimited file which maps the OTU ids to original sequence ids:

DENOVO1 IS0AYJS04JQKIS;sample=Mw_01
DENOVO2 IS0AYJS04JL6RS;sample=Mw_01
DENOVO3 IS0AYJS04H4XNN;sample=Mw_01
...
hits.txt

TAB-separated file, three-columns, where each column contains: the matching sequence, the representative (seed) and the identity (if available, see Definition of identity):

IS0AYJS04JE658;sample=Mw_01; IS0AYJS04I4XYN;sample=Mw_01 99.4
IS0AYJS04JPH34;sample=Mw_01; IS0AYJS04JVUBC;sample=Mw_01 98.0
IS0AYJS04I67XN;sample=Mw_01; IS0AYJS04JVUBC;sample=Mw_01 99.7
...
otuschim.fasta
(only for ‘denovo_greedy’, ‘denovo_swarm’ and ‘open_ref’ mathods, when -c/--rmchim is specified) FASTA file containing the chimeric otus.

Warning

Trimming the sequences to a fixed position before clustering is strongly recommended when they cover partial amplicons or if quality deteriorates towards the end (common when you have long amplicons and single-end sequencing), see Quality filtering.

Note

De novo OTUs are renamed to DENOVO[N] and reference OTUs to REF[N].