Overview

Welcome to the User Manual of FooDMe. This documentation will guide you through your first steps with the pipeline.

Aim

The aim of this workflow is the taxonomic assignement of paired-end Illumina amplicon sequencing reads. It was developped with a focus of food and feed authenticity control.

Workflow description

The workflow consists of three main steps that will be performed sequentially: reads pre-processing, clustering, and taxonomic assignment.

Input data

The pipeline expects paired-end Illumina reads, provided as paths in a tabular file. See the configuration help for more details.

Pre-processing

As a first analysis step, primers will be trimmed from the reads. By default primers are only matched on the 5' end of the reads. In some cases (e.g. Sequencing is longer than the amplicon length) one may want to trim primers on the 3' end as well. This behaviour is supported and can be enabled in the parameters.

The reads will then be pre-processed for quality trimming on the 3' end using a sliding window checking for minimal quality requirements.

Clustering

FooDMe implements three different clustering strategies to choose from. Each has specific performances that can be better suited to your specific needs.

Identity clustering

For this strategy, sequences are first dereplicated and ranked by abundance. Each sequence is then compaired to an itinially empty list of centroids and, based on sequence similarity, is either assigned as a new centroid or merged with the most similar existing centroid.

The degree of identity required for clustering the sequences can be freely set, 0.97 being a commonly accepted value and 1.0 corresponding to a dereplication.

This approach is very good at smoothing out seqencing noise but can also result in the clustering of highly similar sequences for different organisms.

Dereplication

With this strategy, only identical reads will be clustered together. This is effectively implemented as an identity clustering strategy with 100% identity. This is more suitable for high sensitivity identification of amplicons but the sequencing noise will not be filtered.

Due to the large number of clusters resulting from this strategy this is also the most computationally expensive of the three.

Denoising

For the denoising strategy, the nucleotide substitution rate of the sequencing run will be modelled based on the available data. Using this model, reads can be corrected for sequencing-induced substitutions. This results in a typically low number of cluster, close to the biological reality of the sample.

Note

Clusters are often refered to as Operational Taxonomic Units (OTUs) for dereplication and identity clustering and as Amplicon Sequence Variants (ASVs) for denoising.

Taxonomic assignment

BLAST search and filtering

The representative sequences for each cluster are compaired to a user-provided nucleotide database using Basic Local Alignment Search (BLAST) and references satisfying specified similarity critera are recovered.

In most cases, only part of the database is relevant for each application. For this purpose it is possible to specify a taxonomic node (for example Vertebrates) to which to limit the BLAST search.

Specific taxa can also be irrelevant, such as common contaminants or extinct species. This taxa can be provided as a list of identifiers in a text file and will be filtered out of the BLAST search.

Info

We use the NCBI Taxonomy nomenclature, including lineages and identifiers. See https://www.ncbi.nlm.nih.gov/taxonomy for more details.

Because this typically results in a large number of matching results (and taxa), the matches can be post-filtered based on their alignment quality as measured by the alignement bit-scores.

Taxonomic consensus

As this process often results in a unclear mix of taxa, a consensus can be determined based on the underlying taxonomic hierachy and a minimal agreement level that can be freely set between strict majority (0.51) and last-common ancestor (1.0).

Ouput

Statistics on the results of each processing steps as well as the final results with the composition of each samples are saved in tabular files and in an HTML report whose tables can be exported to popular formats such as Excel or PDF. More info here.

Workflow chart

Below is a schematical overview of the data processing:

flowchart TD
    A[Raw data] --> B(Primer trimming)
    subgraph Pre-processing
    B ---> C(Quality trimming)
    end
    C --->|OTU| D1(Read merging)
    C --->|ASV| D2(Denoising)
    subgraph Clustering
    D1 ---> E1(Cluster filtering)
    D2  ---> E2(Read merging)
    end
    Z1(Branch filter) ---> Z0[(BLAST database)]
    Z2(Blocklist) ---> Z0
    Z0 ---> F(Local Alignment search)
    E1 ---> F
    E2 ---> F
    Y[(Taxonomy definitions)] ---> H(Consensus determination)
    subgraph Taxonomic assignement
    F ---> J(Accession filtering)
    J ---> G(Bitscore filtering)
    G ---> H
    end
    H ---> I[Report]