1 of 34

Taxonomic Assignment

2 of 34

  • Your representative sequence (the file that you have obtained from DADA2)
  • A Reference Database
  • A pre-trained classifier

What you need for taxonomic assignment

3 of 34

Taxonomy classification workflow

4 of 34

Compare observed sequences to annotated reference sequences to make taxonomic assignments

NOTE: This process is the same regardless of which algorithm you choose.

The goal

5 of 34

Taxonomy classification workflow

6 of 34

Your query

FASTA file -- your output from QC step --enclose in the rep-seqs_16S.qza file

7 of 34

Taxonomy classification workflow

8 of 34

Taxonomic assignment of observed sequences

Reference sequences

9 of 34

Taxonomy classification workflow

10 of 34

Taxonomic assignment of observed sequences

Reference sequences

Map taxonomy

11 of 34

Taxonomy classification workflow

12 of 34

Taxonomic assignment of observed sequences

The output --> taxonomy.qza

Feature

Taxon

Confidence

13 of 34

RESCRIPt

REference Sequence annotation and CuRatIon Pipeline RESCRIPt is a QIIME 2 plugin to support a variety of operations for managing and curating reference sequence databases, DNA/RNA sequence data, and taxonomic data.

https://library.qiime2.org/plugins/rescript/27/

14 of 34

The algorithms

15 of 34

BLAST -Local Alignment

BLAST is an acronym for Basic Local Alignment Search Tool. BLAST represents both an algorithm and a suite of tools that implement said algorithm. The primary purpose of BLAST is to search a collection of target sequences for alignments that match a query sequence.

BLAST can be run as a web interface from NCBI and a standalone downloadable tool.

16 of 34

BLAST -Local Alignment

Local alignments are used when we need to find the region of maximal similarity between two sequences. When performing local alignments, the algorithms look for the highest-scoring (partial) interval between the two sequences

BLAST search results are sequences that have similarities to the query sequence. As the name says, the results of a BLAST search are local alignments; thus, a blast result is generally a partial match of the query sequence to a target sequence in the database.

17 of 34

BLAST -Local Alignment

E-values were designed to bring some level of confidence to the search results and are defined as the number of hits one can “expect” to see by chance when searching a database of a particular size.

A smaller the e-value is “better”.

18 of 34

Problem using BLAST

BLAST is not an optimal aligner. This means that it may not find all alignments and there are limits for how short and how long sequences may be for BLAST to operate correctly. Also, BLAST is tuned for performance and efficiency; its primary purpose is to search a large body of known information for similarities (hits).

Against Local Database

  • If your query sequence has repetitive elements (also called low complexity regions), BLAST automatically and silently filters out hits to these

Against NCBI

  • The taxonomy names are up-to-date and correct, but the decision to which taxa the sequence belongs is a decision of who submits it.
  • See point before

19 of 34

Vsearch - Global Alignment

In global alignments, the bases of both sequences are arranged next to one another over their entire length. Each base of the first sequence is matched to another base or a “gap” of the second sequence.

  • We use global alignments when we need to look for the similarities over the entire length of both sequences.
  • We use global alignments when we are looking for an arrangment that maximizes the similarities over the entire length of both sequences:

THISLINE - ISALIGNED

Change gap penality

Underperforms when the amount of data is very high

20 of 34

Sklearn - Classify reads by taxon using a fitted classifier�

21 of 34

Where does the machine learning classifier come from?

Using the reference database (our examples) with train a classifier (create), like said:

“if you see a sequence like this one, you should be outputting a taxonomy string like this one”

Sklearn - Classify reads by taxon using a fitted classifier�

22 of 34

Sklearn - Classify reads by taxon using a fitted classifier�

Imagine a bag fill of k-mer put in a machine learning classifier

7-mer

23 of 34

Sklearn - Classify reads by taxon using a fitted classifier�

Naive Bayes (posterior probability) algorithm for multinomial distributed data is used to create a train set (learn from a bag of k-mers to taxonomic string)

It is a classification technique based on Bayes’ Theorem with an assumption of independence among predictors. In simple terms, a Naive Bayes classifier assumes that the presence of a particular feature in a class is unrelated to the presence of any other feature.

For example, a fruit may be considered an apple if it is red, round, and about 3 inches in diameter. Even if these features depend on each other or upon the existence of the other features, all of these properties independently contribute to the probability that this fruit is an apple, which is why it is known as ‘Naive’.

24 of 34

Reference Databases

25 of 34

SILVA

The SILVA database is a large project developed and maintained by the Leibniz Institute DSMZ-German Collection of Microorganisms and Cell Cultures in Bremen, Germany, in cooperation with a company.

SILVA is an interdisciplinary project of biologists and computer scientists to provide:

  • fully aligned and up to date small (16S/18S, SSU) and large (23S/28S, LSU) subunit ribosomal RNA "Parc" databases on the webpage as well as ARB files
  • preconfigured subsets of only high quality, full-length sequences as ARB & FASTA files (SSU/LSU Ref)

26 of 34

SILVA

27 of 34

SILVA

28 of 34

GTDB

The Genome Taxonomy Database (GTDB) is an initiative to establish a standardized microbial taxonomy based on phylogenomics.

https://gtdb.ecogenomic.org/about

29 of 34

UNITE

Among the databases dedicated to ITS sequences for the identification of fungi, a central place is taken by UNITE (User-friendly Nordic ITS Ectomycorrhiza Database).

UNITE makes available the collection of only high-quality fungal ITS sequences stored in the International Nucleotide Sequence Databases (INSDC) and well - curated taxonomic and ecological annotations from a third-party distributed effort.

In particular, the taxonomic annotation is improved following the Index Fungorum (http://www.indexfungorum.org/) classification scheme.

30 of 34

UNITE

31 of 34

UNITE

32 of 34

PR2

The Protist Ribosomal Reference Database 's aim is to provide a reference database of carefully annotated 18S rRNA gene sequences using nine unique taxonomic fields (from domain to species).

At present, it contains over 220,000 sequences. Although it focuses on protists, it also contains sequences from metazoa, fungi, and plants and a limited set of 16S sequences from plastids and bacteria.

Several metadata fields are available for many sequences, including geo-localization, whether it originates from a culture or a natural sample, host type, etc. … Experts from each taxonomic group annotate PR2.

33 of 34

PR2

https://pr2-database.org/

34 of 34

PR2