---
title: "DNABarcodes"
output:
BiocStyle::html_document:
toc: true
BiocStyle::pdf_document:
toc: true
---
```{r style, echo = FALSE, results = 'asis', warnings=FALSE, messages=FALSE}
BiocStyle::markdown()
```
```{r setup, include=FALSE}
knitr::opts_chunk$set(cache=FALSE)
```
# DNABarcodes
The `DNABarcodes` package offers a function to create DNA barcode sets capable
of correcting substitution errors or insertion, deletion, and substitution
errors. Existing barcodes can be analysed regarding their minimal, maximal and
average distances between barcodes. Finally, reads that start with a (possibly
mutated) barcode can be demultiplexed, i.e., assigned to their original
reference barcode.
The most common use cases are:
* `create.dnabarcodes` is used to create sets of DNA barcodes with error correction properties.
* `demultiplex` is used to demultiplex a set of reads based on the used set of DNA barcodes
* `analyse.barcodes` is used to analyse the properties of an existing set of DNA barcodes
and to assess its error correction capabilities
# Using DNABarcodes
## Creating a Pristine Set of DNA Barcodes
Suppose we want to generate a set of 5bp long barcodes with default settings.
The default enforces a minimum Hamming distance of 3 between each DNA barcode,
which is sufficient to correct at least one substitution error:
```{r eval=TRUE, collapse=TRUE}
library("DNABarcodes")
mySet <- create.dnabarcodes(5)
show(mySet)
```
In default mode, the Conway lexicographic algorithm is used to generate the
set. Conway's algorithm is simple and most of the time efficient enough. When
sufficient computing power is available, Ashlock's evolutionary algorithm
should be used. The parameter to change the set generation algorithm is named
*heuristic*.
```{r eval=TRUE, collapse=TRUE}
mySetAshlock <- create.dnabarcodes(5, heuristic="ashlock")
show(mySetAshlock)
```
Importantly, the set of DNA barcodes is most often larger when Ashlock's
algorithm is used.
### Sets with Larger Number of Correctable Errors
Finally, we want to create a set of 5bp long DNA barcodes that support the
correction of 2 substitutions. For that, we enforce a minimum distance of 5
using the parameter *dist*.
```{r eval=TRUE, collapse=TRUE}
mySetDist5 <- create.dnabarcodes(5, dist=5, heuristic="ashlock")
show(mySetDist5)
```
The number of errors $k_c$ that a set of DNA barcodes can correct is expressed
by the following formula. The variable $dist$ gives the minimum distance
of the set.
$$k_c \leq \left\lfloor \frac{dist - 1}{2} \right\rfloor$$
The number of $k_d$ of detectable errors is given as:
$$k_d \leq dist - 1$$
The following table shows common distances $dist$ and their error correction
($k_c$) and detection ($k_d$) properties:
$dist$ $k_c$ $k_d$
------ ----- -----
3 1 2
4 1 3
5 2 4
6 2 5
7 3 6
8 3 7
9 4 8
To obtain a large enough set for the targeted number of samples, it is often
necessary to increase barcode length. Here, we want to correct 2 substitution
errors and want to target at least 20 samples:
```{r eval=TRUE, collapse=TRUE}
show(length(create.dnabarcodes(5, dist=5, heuristic="ashlock")))
show(length(create.dnabarcodes(6, dist=5, heuristic="ashlock")))
show(length(create.dnabarcodes(7, dist=5, heuristic="ashlock")))
```
Therefore, 7bp long DNA barcodes should be used.
### Sets of DNA Barcodes Capable of Correcting Insertions, Deletions, and Substitutions
To generate sets of DNA barcodes that support the correction of insertions,
deletions, and substitutions (e.g., for the PacBio platform), a different
distance metric must be used. In a DNA context (i.e., the DNA barcode is
surrounded by other DNA nucleotides), the Sequence-Levenshtein distance is the
right choice. Here, we generate a set of 5bp long DNA barcodes capable of
correcting one indel or substitution error. The parameter to change the
distance metric is named *metric*:
```{r eval=TRUE, collapse=TRUE}
mySeqlevSet <- create.dnabarcodes(5, metric="seqlev", heuristic="ashlock")
show(mySeqlevSet)
```
The requirements of the Sequence-Levenshtein distance are more stringent than
those of the Hamming distance. The number of DNA barcodes in the set is
therefore smaller, but the set is more robust.
### Applying Different Filters
By default, `create.dnabarcodes` filters all those DNA sequences that contain
triplets, show a GC bias, or are self-complementary. Some of these filters may
be unnecessary on current or future platforms. For example, with the Illumina's
Sequencing by Synthesis technology, the triplet filter is unnecessary. Here, we
generate a default set of DNA barcodes of length 5bp without filtering triplets:
```{r eval=TRUE, collapse=TRUE}
mySetTriplets <- create.dnabarcodes(5, heuristic="ashlock", filter.triplets=FALSE)
show(mySetTriplets)
```
Note that the size of the pool of candidate barcodes is larger (640) than for
the default setting that includes the triplet filter (592). Potentially, this
allows the creation of larger sets, which we observed for longer DNA barcodes.
## Subsetting an Existing Set of DNA Barcodes
In the following, we describe methods to generate subsets of existing sets of
DNA barcodes. Often, a researcher already has a pre-existing set of DNA
barcodes, for example in chemical form as indexes from his supplier. When a
smaller number of samples needs to multiplexed, he can choose a more robust
subset of the existing indexes to get better error correction capabilities.
The set of candidate barcodes is given to `create.dnabarcodes` through the
parameter *pool*.
For example, the researcher may have a RNASeq library preparation kit that
includes 48 indexes. The set can already be used to correct 1 substitution. He
creates a robust subset for the correction of 2 substitutions the following
way:
```{r eval=TRUE, collapse=TRUE}
data(supplierSet)
myRobustSet <- create.dnabarcodes(7, dist=5, pool=supplierSet, heuristic="ashlock")
show(myRobustSet)
```
Alternatively, we may want to create a subset that is capable of correcting
indels in addition to substitutions:
```{r eval=TRUE, collapse=TRUE}
myRobustSetSeqlev <- create.dnabarcodes(7, metric="seqlev", pool=supplierSet, heuristic="ashlock")
show(myRobustSetSeqlev)
```
## Demultiplexing
Demultiplexing is processing step where reads are assigned to their samples.
Demultiplexing is achieved easily with the `demultiplex` function. In the
following example we assume to have a file that contains all reads that start
with a DNA barcode. The used set contains 48 7nt long DNA barcodes and was
generated to correct up to one substitution.
```{r eval=TRUE, collapse=TRUE}
data(mutatedReads)
demultiplex(head(mutatedReads), supplierSet)
```
It is important to supply the correct distance metric to `demultiplex` using
the parameter *metric*. Otherwise the result will change unintentionally and in
unexpected ways.
## Analysing a Set of DNA Barcodes
Suppose we obtained a set of DNA barcodes in form of sample indexes from our
library preparation supplier. We want to analyse the set to understand the
errors that can be corrected or detected. The function `analyse.barcodes`
easily does just that:
```{r eval=TRUE,collapse=TRUE}
analyse.barcodes(supplierSet)
```
The output table lists mean, median, minimum, and maximum distances of each
pair of DNA barcodes in the set. Finally, it lists the guaranteed error
correction and detection capabilities that can be reached for the set. The
analysis is presented for a choice of distance metrics: Hamming,
Sequence-Levenshtein, and Levenshtein.
# Distance Metrics
The `DNABarcodes` package currently supports four distance metrics. Their
capabilities and properties are as follows:
A high enough Hamming Distance (`metric = "hamming"`) allows the
correction/detection of substitutions. Due to the ignorance of insertions and
deletions, any changes to the length of the barcode as well as DNA context are
ignored, which makes the Hamming distance a simple choice.
A high enough Sequence Levenshtein distance (`metric = "seqlev"`) allows the
correction/detection of insertions, deletions, and substitutions in scenarios
where the barcode was attached to a DNA sequence. Your sequence read, coming
from a Illumina, Roche, PacBio NGS machine of your choice, should then start
with that barcode, followed by the insert or an adapter or some random base
calls.
A high enough Levenshtein distance (`metric = "levenshtein"`) allows the
correction/detection of insertions, deletions, and substitutions in scenarios
where the barcode was not attached anywhere, respective where the exact
beginning and end of the barcode is known. This is as far as we know in no
current NGS technology the case. Do not use this distance metric, except you
know what you are doing.
A high enough Phaseshift distance (`metric = "phaseshift"`) allows the
correction/detection of substitutions as well as insertions or deletions that
occurred to the front of the DNA barcode. The intended target technology of
this distance is the Illumina Sequencing by Synthesis platform. We consider
this metric to be highly experimental.
# Set Generation Heuristics
We support four different algorithms for the generation of sets of DNA
barcodes. Each has its particular advantages and disadvantages.
The heuristics `conway` and `clique` produce fast results but are not nearly as
good as the heuristics `sampling` and `ashlock`. The `clique` heuristic is
marginally slower and needs more memory than the `conway` heuristic because it
first constructs a graph representation of the pool.
The heuristic `ashlock` is assumed to produce the best heuristic results with a
reasonable parameter configuration. New users should first try `conway` and
then `ashlock` with default parameters.
## Configuration of `Ashlock` Heuristic
The Ashlock heuristic is an evolutionary algorithm. Briefly, it iterates over a
*population* of so-called chromosomes (not to be confused with actual genomic
chromosomes), while trying to improve set creation results in each step. In
each iteration, it retains the best chromosomes and replaces the rest with
mutated copies of the best ones.
The algorithm has two parameters:
* The number of iterations (`iterations`)
* The size of the population (`population`)
More iterations and a larger population will (possibly) increase the size of
the resulting set of DNA Barcodes but will also increase running time
considerably.
We believe that 100 is a very good default for the number of iterations. The
best chance, in our opinion, is to increase the number of chromosomes in the
population considerably, for example up to 500 or even 1000.
The following examples show the increase in set size:
```{r eval=FALSE, collapse=TRUE}
length(create.dnabarcodes(10, metric="seqlev", heuristic="ashlock",cores=32))
## 1) Creating pool ... of size 488944
## 2) Initiating Chromosomes done
## 3) Running Greedy Evolutionary done
## 2126
length(create.dnabarcodes(10, metric="seqlev", heuristic="ashlock",cores=32, population=500, iterations=250))
## 1) Creating pool ... of size 488944
## 2) Initiating Chromosomes done
## 3) Running Greedy Evolutionary done
## 2133
```