A set of scripts to cluster mutational trajectories into genotypes and cluster genotypes by background

# Installation

These scripts are available on pypi and can be installed with

pip install muller


To update the scripts to the newest version, simply run

pip install muller --upgrade


Then run the scripts using

lineage [args]


It is also possible to simply clone the package, although the additional required packages would then need to be installed separately.

git clone https://github.com/cdeitrick/muller_diagrams.git
cd muller_diagrams
lineage [args]


# Requirements

The scripts require a few python packages to work. Each of these can be installed using pip install [package] or conda install [package].

• dataclasses (if using a python version below 3.7)
• loguru
• matplotlib
• pandas
• pygraphviz
• scipy
• seaborn
• xlrd (to read excel files)

If the package pygraphviz throws an error during installation, it is usually because it can't find the correct dependancies in the current environment. Install the dependancies using

Linux:

sudo apt-get install python-dev graphviz libgraphviz-dev pkg-config


Mac:

brew install graphviz


or the equivalent package manager on your system.

## Optional Packages

• tqdm: If tqdm is also installed, the scripts will display a progressbar for large datasets.
• beautifulsoup4: Sometimes the encoding of csv files is ambiguous (the scripts throw a UnicodeDecodeError). If beautifulsoup4 is installed the scripts will attempt to correct encoding errors.

# Sample Usage

The scripts currently default to hierarchical clustering using the binomial distance. More information is available in the "description" folder. Use python to call the "muller" folder:

Muller.py --input [input filename] --output [output folder]


Run with default parameters.

Muller.py --input [filename] --frequencies 0.05 --detected 0.10


Groups genotypes in groups of 0.05 (i.e. [0.00, 0.05, 0.10, ... , 0.90, 0.95, 1.00]) based on each genotype's maximum frequency. Each genotype in each group is then sorted by the timepoint it was first detected (the first timepoint where the frequency was greater than 0.10). Output files are saved to the same folder as the input table.

# General Workflow

Flowcharts for each individual step can be found under docs/flowcharts.

# Script Options

## General Options

-h, --help
Show a help message and exit
--name
Prefix to use when naming the output files. defaults to the dataset filename.
-i, --input
The table of trajectories to cluster. Must be an excel file or csv/tsv file.
The delimiter will be inferred from the file extension.
-o,  --output
The output folder to save the files to.
The number of processes to use. This is only relevant for very large datasets.
-d, --detection
The uncertainty to apply when performing
frequency-based calculations. For
example, a frequency at a given timepoint
is considered undetected if it falls
below 0 + detection.
--fixed
The minimum frequency at which to
consider a mutation fixed. Defaults to
1 - uncertainty
-s, --significant
[0.15] The frequency at which to consider a genotype
significantly greater than zero.
-f, --frequencies
[0.10] The frequency cutoff or step to use when sorting genotypes.
May be a comma-separated string of frequencies, or a set inverval
to use when generating the frequency breakpoints. This affects
the filtering step and the nesting step.
For example, a value of 0.15 will use the frequencies 0,.15,.30,.45...
--genotypes                 Indicates that the input table contains genotypes rather
than mutational trajectories. This will skip the filtering and clustering steps.
--sheetname
Specifies the sheet to use when the input is an excel file. Defaults to
the first sheet in the spreadsheet.
--gene-alias ALIAS_FILENAME
An optional two-column file with more accurate gene
names. This is useful when using a reference
annotated via prokka.


## Filtering Options

--dasable-all-filters
Disables the genotype filtering step.
--disable-filter-single
Keep trajectories only detected at a single timepoint.
--disable-filter-startsfixed
Keeps mutational trajectories which begin the experiment fixed.
--filter-constant
[0.10] Sets the delta value by which a mutational trajectory must vary by to not
be removed for being a constant mutation. Set to 0 to disable


## Clustering Options

-m, --method                Selects the clustering method to use. 'two-step' will use the original two-step
method of sorting trajectories into genotypes while 'hierarchy' will use
hierarchical clustering to do the clustering. Defaults to 'matlab'
--metric                    Used to select the distance metric when --method is set to 'hierarchy'.
Available Options:
'similarity', 'binomial' [Default] Uses the binomial test implemented in the original matlab scripts as a distance metric.
'jaccard'               Uses the Jaccard distance between two series to determine the distance metric.
'minkowski'             Uses the minkowski distance as a distance metric. Primarily influenced by the
difference between two series.
'pearson'               Uses the pearson correlation coefficient as the distance metric. Primarily
influenced by the correlation of two series against each other.
'combined'              A combination of the 'pearson' and 'minkowski' distances to account for the
correlation of two series as well as the difference between them.
-r --similarity-cutoff
[0.05] Used when grouping trajectories into genotypes.
Maximum p-value difference to consider trajectories related when using
the two-step method, and selects the maximum distance to consider
trajectories related when --method is hierarchy.
-d, --difference-cutoff     [0.10] Only used when --method is twostep.
Used to unlink unrelated trajectories present in a genotype. Is not used
when using hierarchical clustering.
-g, --known-genotypes
Path to a file listing trajectories which are known to be in the same genotype.
Each line in the file represents a single genotype, and each line should be a
comma-separated list of trajectory labels as they appear in the input dataset.
--filename-pairwise
A table of pairwise distance calculations computed in a previous run using
identical input parameters. This is only usefull if the dataset being re-run
would other wise take a very onlg time to process (such as data from the
Long Term Evolution Experiment).


## Nesting Options

--additive
[0.03] Controls how the additive score between a nested and
unnested genotype is calculated. Defaults to the
detection cutoff value.
--subtractive
Controls when the combined frequencies of a nested and
unnested genotype are considered consistently larger
than the fixed cutoff.Defaults to the detection cutoff
value. (default: None)
--derivative
Controls how much a nested and unnested genotype
should be correlated/anticorrelated to be considered
significant (default: 0.01). Correlation implies a positive relationship
between the nested/unnested genotypes while anticorrelation is evidence
against nesting the unnested genotype under the nested genotype.
--known-ancestry
A tab-delimited file designating the known ancestry of certain
genotypes. The left column should be the genotype to nest,
right column should be its parent. Column names are ignored.
Genotype names are generated during the clustering step,
so this is only useful when re-running the analysis.


## Graphics Options

--genotype-colors           Path to a file with a custom genotype colorscheme. The file should be tab-delimited
with a genotype name (ex. 'genotype-13') in the first column and a HEX color code
(ex. '#F5674A') in the second. These colors will override the default colorscheme.
--no-outline
Disables the white ouline surrounding each series in the muller plots.
--no-render
Disables .svg renders.
--highlight
A comma-separated list of genotype names or annotations to highlight in the generated graphics.
--highlight-color
[#F34A20] What the color of highlighted genotypes should be. Only HEX color codes are supported.


# Input Dataset

The script operates on a table listing all mutations and their corresponding frequencies at each timepoint (refered to as "trajectories" in this script) or a table with each genotype and frequency at each timepoint (ex. the genotype table in the examples folder). The table must have a column named Trajectory with labels for each mutational trajectory (or Genotype when using --genotype) and integer columns for each timepoint. The labels are solely used to identify trajectories belonging to a specific genotype, and must be integers. All other columns will be ignored when calculating genotypes and genotype clusters. The frequencies can be represented as either a number between 0 - 1, a number between 0 - 100 or as percentage. The Trajectory and Genotype columns can contain any kind of label, but must be unique for each trajectory/genotype.

Population Trajectory Chromosome Position Class Mutation 0 17 25 44 66 75 90
B2 1 1 38102 SNP C>T 0 0 26.1% 100% 100% 100% 100%
B2 2 1 62997 SNP T>G 0 0 0 52.5% 45.4% 91.1% 91%
B2 3 1 78671 SNP A>C 0 0 0 14.7% 45% 92.4% 88.7%
B2 4 1 96585 SNP T>G 0 0 0 0 21.1% 81.1% 81.3%
B2 5 1 115010 SNP G>T 0 0 0 40.3% 48.9% 5.7% 8%
B2 t16 1 299332 SNP C>T 0 0 0 0 20.9% 20.9% 0
B2 6 1 156783 SNP C>G 0 0 0 0 0 100% 100%
B2 7 1 176231 SNP T>A 0 0 0 27.3% 78.1% 100% 100%
B2 8 1 205211 SNP C>T 0 0 0 0 34.5% 83.3% 79.3%
B2 9 1 223199 SNP C>G 0 0 0 0 0 26.9% 34%
B2 trajectory-10 1 262747 SNP T>C 0 0 11.7% 0 0 0 10.3%
B2 trajectory-11 1 264821 SNP C>T 0 0 0 10.8% 15.1% 0 0
B2 trajectory-12 1 298548 SNP G>A 0 12.5% 0 15.3% 18.1% 17.5% 19.1%
B2 trajectory-13 1 299331 SNP G>A 0 0 0 0 25.8% 5.7% 7.5%
B2 trajectory-14 1 299332 SNP C>T 0 38% 43.2% 0 0 0 0
B2 t15 1 299332 SNP C>T 0 0 6.6% 10.4% 6.2% 0 0
B2 t16 1 299332 SNP C>T 0 0 0 0 20.9% 20.9% 0
B2 t17 1 299332 SNP C>T 0 0 0 0 0 26.6% 31.2%
B2 t18 1 299332 SNP C>T 0 0 0 11.5% 0 13.1% 0
B2 t19 1 299332 SNP C>T 0 0 0 18.8% 17.1% 23.2% 24.4%
B2 20 1 299332 SNP C>T 0 0 0 13.8% 29.5% 0 8.1%
B2 21 1 299332 SNP C>T 0 0 0 11.4% 0 11% 12.3%

# Output

All files are prefixed by the name of the original input table if the --name parameter is unfilled.

## Tables

### Timeseries tables

• .muller_genotypes.tsv
• .muller.trajectories.tsv
• tables/.muller_genotypes.original.tsv
• tables/.trajectories.original.tsv

Tables listing the genotypes and trajectories encountered in the analysis. The trajectory tables also link each trajectory to its respective genotype. There are two versions of these tables: one set with the original input trajectories and the initial calculated genotypes and another set with the final trajectories and genotypes left in the analysis after the filtering step. The trajectory tables include all columns from the input trajectory table as well as the timeseries and annotation columns used in the analysis.

Example Genotype Table:

Genotype 0.000 17.000 25.000 44.000 66.000 75.000 90.000
genotype-1 0.000 0.380 0.432 0.000 0.000 0.000 0.000
genotype-2 0.000 0.000 0.000 0.403 0.489 0.057 0.080
genotype-3 0.000 0.000 0.000 0.000 0.000 1.000 1.000
genotype-4 0.000 0.000 0.261 1.000 1.000 1.000 1.000
genotype-5 0.000 0.000 0.000 0.273 0.781 1.000 1.000
genotype-6 0.000 0.000 0.092 0.052 0.031 0.000 0.052
genotype-7 0.000 0.000 0.000 0.000 0.278 0.822 0.803
genotype-8 0.000 0.000 0.000 0.336 0.452 0.918 0.899
genotype-9 0.000 0.000 0.000 0.076 0.043 0.219 0.255
genotype-10 0.000 0.021 0.000 0.086 0.182 0.095 0.058

### Tables for ggmuller

• tables/.ggmuller.populations.tsv
• tables/.ggmuller.edges.tsv

These tables are designed for use with the ggmuller r package. The populations table describes the population/abundance of each genotype at each timepoint while the edges table describes the ancestry relationship between genotypes.

This table is generated using the scipy python package. It describes the agglomeration of clusters starting with the individual trajectories, as well as the mean, variance, and trajectory count of each cluster. Columns:

• left, right: The two sub-clusters merged to create the current clusters
• clusterId: The id assigned to this cluster. Note that since the individual genotypes are not included in the table, the clusters are numbered in order starting with 1 + the total number of genotypes.
• distance: The distance between the two sub-clusters.
• observations: The number of mutational trajectories contained in this cluster.

left right distance observations resultingCluster
7 18 0.034 2 19
13 17 0.175 2 20
8 11 0.199 2 21
2 5 0.239 2 22
10 3 0.279 2 23
9 12 0.370 2 24
23 6 0.529 3 25
22 21 0.624 4 26
26 1 0.708 5 27
24 16 0.760 3 28
14 25 0.786 4 29
15 20 0.988 3 30
29 27 1.094 9 31
31 19 1.358 11 32
30 28 1.362 6 33
4 32 1.499 12 34
33 0 2.125 7 35
34 35 4.943 19 36

### Distance Matrix

• tables/.distance.tsv

A table of pairwise distance values between each trajectory.

### Muller table

• tables/.muller.tsv

The converted form of the .ggmuller.populations.tsv and .ggmuller.edges.tsv used to generate the muller plots. This file is created from the r script, described later.

## Graphics

Each of the output plots use the same palette for genotypes and trajectories. A genotype colored a shade of blue will share that color across all graphs and diagrams which depict that genotype. There are two palettes: one to indicate each clade in the geneology and one to easily distinguish between different genotypes. Each graphic is created with both palettes, and some are provided in multiple formats for convienience.

### Muller Plots

• .muller.annotated.png
• graphics/distinctive/.muller.annotated.distinctive.png
• graphics/distinctive/.muller.annotated.distinctive.svg

The main value of a muller plot is to quickly visualize abundance and geneology of genotypes over the course of an evolution experiment.

### Lineage Diagrams

• .lineage.png
• graphics/.lineage.distinctive.png

These are simple flowcharts indicating the relationship between genotypes and clades. The original genotype of each clade are shown to arise in "genotype-0", the root background. The ancestry of all other genotypes are then shown relative to these clades.

### Timeseries plots

• .genotypes.png
• .genotypes.filtered.png
• .trajectories.distinctive.png

Timeseries plots of the frequency of each trajectory and genotype at each timepoint. Trajectories are colored according to which genotype they were grouped into. The .genotypes.filtered.png file includes trajectories that were filtered out during the filtering step (clored black).

### Distance Heatmap

• graphics/.heatmap.distance.png

A pairwise comparison of the calculated distance between each mutational trajectory. Trajectories are grouped by the final genotype. The heatmap will be annotated with the distance values if there are fewer than thirty total trajectories in the analysis.

### Dendrogram

• graphics/.dendrogram.png Shows the arrangement and distance between clusters and member trajectories. Not available with --method twostep.

## Scripts

• scripts/example.r

One external script is used during the course of this analysis. The r script is based on the ggmuller package implemented in r, and is used to convert the genotypes data into a format required to generate the muller plots. This script also generates a basic muller plot (/graphics/distinctive/.muller.png), although all other muller plots are created with the python implementation.

## Supplementary files

• supplementary-files/.json

A json-formatted file with all parameters used in the analysis.

• supplementary-files/.nestscores.tsv

Lists the scores between each genotype and the corresponding candidate ancestry genotypes. The highest score above or equal to 1 determines the parent genotype.

## Project details

Uploaded source
Uploaded py3