ali-nayeem / mammle

Phylogeny estimation based on multiobjective application-aware MUSCLE and maximum likelihood ensemble

MAMMLE: A framework for phylogeny estimation based on multiobjective application-aware MSA and maximum likelihood ensemble

MAMMLE is a software framework for inferring better phylogenetic trees from unaligned sequences by hybridizing MUSCLE with multiobjective optimization strategy and leveraging multiple Maximum Likelihood hypotheses. MAMMLE may offer a significant improvement (upto 27% in our experiments on BAliBASE 3.0) in tree accuracy over MUSCLE.

Fig.1: Simplified phylogenetic reconstruction pipeline of MAMMLE framework. The components open to modification are marked with a blue shade.

Multiobjective Application-aware MUSCLE

For MAMMLE, we develop Multiobjective Application-aware MUSCLE by embedding the following four objective functions, identified based on their better correlation to the tree accuracy, within the iterative phase of MUSCLE:

1. Maximize similarity for columns containing gaps (SIMG)
2. Maximize similarity for columns containing no gaps (SIMNG)
3. Maximize sum-of-pairs (SP)
4. Minimize number of gaps (GAP)

Fig.2: High-level workflow of multiobjective application-aware MUSCLE for a single weight vector. Steps (3.4 to 3.6) added/modified on the original MUSCLE are marked with red color. This figure is modification of the original image taken from https://doi.org/10.1093/nar/gkh340.

Software Commands

Below we provide the software commands used for our experimentation. Please note the provided commands of MUSCLE and RAxML only represent the current components of MAMMLE framework. One can easily replace these components with other appropriate tools/commands which is the main strength of this framework.

• MUSCLE (version 3.8.1551) command

muscle -in [unaligned_sequences] -out  [output_name] -maxiters 20

• RAxML (version 8.2.11) command for reference tree

raxml -f a -m PROTGAMMAAUTO -s [reference_msa]  -n [output_name]  -p 1234 -x 1234 -#100 -T [thread_count]

• RAxML (version 8.2.11) command for ML tree

raxml -m PROTCATWAGF -n [output_name]  -s [estimated_msa]   -p 123456789 -T [thread_count] 	

• MAFFT (version 7.490) command

linsi --auto --amino  --thread [thread_count]  [unaligned_fasta] > [output_name]

• M2Align (version 4.13.1) command

java -cp [path_to_jar] org.uma.khaos.m2align.runner.M2AlignBALIBASERunner [balibase_instance] [path_to_balibase_data] [number_of_evaluations:15000] [population_size:30] [thread_count]

Installation

The current version of MAMMLE has been developed and tested entirely on Linux and MAC.

You need to have:

• Python (version 3.7 or later) with numpy package
• libpython2.7 (install via sudo apt-get install libpython2.7)
• Perl (version 5.15 or later)

Open a terminal and clone latest MAMMLE from our github repository. For example you can use:

git clone https://github.com/ali-nayeem/mammle.git
cd mammle

Give executive permission to the binaries

chmod +x bin/*

Execution

To run MAMMLE using the command-line:

python run_mammle.py -i input_fasta -o output_directory

MAMMLE by default picks the appropriate configurations automatically for you.

Run

python run_mammle.py --help

to see MAMMLE's various options and descriptions of how they work. For example:

  --thread THREAD_COUNT, -t THREAD_COUNT
                        Number of threads (default: 1)
  --input INPUT_SEQUENCES, -i INPUT_SEQUENCES
                        Input unaligned sequences in FASTA format
  --out OUTPUT_DIR, -o OUTPUT_DIR
                        Output tree sequences in NEWICK format
  --weight WEIGHT_VECTORS, -w WEIGHT_VECTORS
                        Input weight vectors in CSV format (default: 30 weight
                        vectors in weight/weights4D-30.csv)