Alignment Filtering

In a typical phylogenomic workflow, you may want to filter problematic alignments before running a phylogenetic analysis. This feature provides multiple ways to filter alignments.

Available filtering methods:

1. Minimal alignment length
2. Data matrix completeness
3. Parsimony informative sites
4. Proportion of missing data

The command is structured as below:

segul align filter <input-option> [alignment-path] <filtering-option> <value>

info

SEGUL will never overwrite your original datasets. When filtering alignments, it will copy the alignments that match the filtering parameters to its output directory. You can also concatenate the resulting alignments instead. This is particularly useful for quickly estimating a species tree based on the filtered alignments.

tip

We recommend starting with the sequence filtering method before using the alignment filtering method. This step will help clean your alignments from problematic sequences, which could result in alignments containing low sample coverage. You can then use the alignment filtering method with the --percent option (see below) to filter out those alignments.

Filtering based on the minimal alignment length

To filter based on the alignment length (also often refer to the number of sites in a textbook or NCHAR in the NEXUS header), we will use the --len option. For example, our directory below contains three alignments:

alignments/
├── locus_1.nexus
├── locus_2.nexus
└── locus_3.nexus

We want to filter an alignment with a minimal length of 500 bp. Use the segul command using the --dir input as below:

segul align filter --dir alignments/ --len 500

We can achieve the same result using the --input or -i option:

segul align filter --input alignments/*.nexus --len 500

Filtering based on data matrix completeness

This filtering method is based on the percentage of minimal taxa present in your alignments, similar to phyluce_align_get_only_loci_with_min_taxa in Phyluce pipeline. For example, given a collection of alignments with 100 taxa with 90 percent completeness, SEGUL will filter alignments containing at least 90 taxa.

We use the --percent option with input in decimals to filter alignments based on data matrix completeness. For example, here we will filter alignments in the alignments directory with 80 percent data matrix completeness:

segul align filter --dir alignments/ --input-format nexus --percent 0.8

You can also write the percentage as .8, and it will parse the same way you write it with 0.

Sometimes, we want to build phylogenetic trees with different data matrix completeness. Using segul, we can filter with multiple data matrix completeness values in a single command. For this task, we use the npercent option:

segul align filter --dir segul --dir alignments/ --input-format nexus --npercent 0.9 0.8 0.75

SEGUL will create an output directory for each data matrix completeness value, with the percentage values suffixed on the directory names. For the example above, it will create three directories named alignments_90p, alignments_80p, and alignments_75p.

Filtering based on parsimony informative sites

SEGUL provides two ways of filtering based on parsimony informative sites:

Based on the minimum input values of informative sites

We use the --pinf option to input the number of informative sites. This option filters alignments that contain at least that number of informative sites.

For example, in the command below, we will filter alignments in the alignments directory with a minimum containing 50 parsimony informative sites:

segul align filter --dir alignments/ --input-format nexus --pinf 50

Based on the percentage of minimal parsimony informative sites

This value is counted based on the highest number of informative sites on your alignments. It is similar to computing data matrix completeness but based on the number of informative sites. You only need to input the percentage value. SEGUL will determine the highest number of informative sites across all input alignments.

For example, the highest number of informative sites in the input alignments is 100. We will filter with a minimal 80 percent of the highest parsimony informative value. Using the --percent-inf .8 option, SEGUL will filter alignments with at least containing 80 parsimony informative sites:

segul align filter --dir alignments/ --percent-inf 0.8

Filtering based on the proportion of missing data

The proportion of missing data is calculated by dividing the total missing data across all sequences in an alignment by the total number of characters. SEGUL will evaluate the proportion of missing data in each alignment and copy the matching alignments to the output directory. To filter the alignments, use the --missing-data option. The value is in decimals.

segul align filter --dir alignments/ --missing-data 0.5

Specifying output directories

By default, SEGUL output directories for filtering are based on the input directory names suffixed with filtering values. The format is as follows:

[input-dir](underscore)[filtering-values]

For example, if your input directory is named alignments/ and the filtering option is a minimal alignment length of 500 bp, the output directory will be alignments_500bp.

The --output or -o option allows you to change the prefix names of the output directory.

Concatenating the results

By default, SEGUL copies the alignments that match the filtering option. Instead of copying the files, you can concatenate them using the --concat flag. All options for the align concat subcommand are available for this task. You are, however, required to specify the --output or -o name and the partition format --part or -p option.

For example, in the command below, we will filter based on alignments with a length of 500 bp and will concat the result:

segul align filter --dir alignments/ --input-format nexus --len 500 --concat --part raxml -output concat_alignment