EmreTasdemirClustalOmega 0.1.0

Pure-Python Clustal Omega multiple sequence alignment implementation

EmreTasdemirClustalOmega

A pure-Python implementation of the Clustal Omega multiple sequence alignment (MSA) algorithm for DNA sequences. No external dependencies required — only the Python standard library.

---

Features

• k-tuple distance matrix — fast pairwise sequence comparison using shared k-mers
• mBed embedding — projects sequences into Euclidean space via reference-sequence distances
• Bisecting k-means clustering — groups sequences before tree construction to reduce complexity
• UPGMA guide tree — builds a hierarchical guide tree both with and without k-means pre-clustering
• Profile HMM alignment — progressive alignment along the guide tree using profile Hidden Markov Models and Viterbi decoding
• Iterative refinement — improves the MSA score via repeated leave-one-out realignment (HHAlign style)
• FASTA input — reads standard .fasta / .fa files; also supports interactive manual input

---

Installation

pip install EmreTasdemirClustalOmega

Requires Python 3.8 or later.

---

Quick Start

As a library

from clustalomega import align

sequences = [
    ("seq1", "ATGCTAGCTAGCT"),
    ("seq2", "ATGCTAGCTAGCC"),
    ("seq3", "ATGCTTGCTAGCT"),
    ("seq4", "TTGCTAGCTATCT"),
]

aligned_blocks, names = align(sequences, k=2)

for name, block in zip(names, aligned_blocks):
    print(f"{name:<10} {block}")

Output:

seq1       ATGCTAGCTAGCT
seq2       ATGCTAGCTAGCC
seq3       ATGCTTGCTAGCT
seq4       TTGCTAGCTATCT

Parameters

Parameter	Type	Default	Description
sequences	list[tuple[str, str]]	—	List of (name, sequence) tuples
k	int	3	k-tuple length for distance calculation
seed	int	42	Random seed for reproducibility
print_ile_yazdirma	bool	False	Print step-by-step output to stdout

Returns: (aligned_blocks, names) where both are lists of strings in the same order.

---

Verbose mode

aligned_blocks, names = align(sequences, k=2, print_ile_yazdirma=True)

This prints the full pipeline output: distance matrix, embedding vectors, clustering steps, guide tree, initial alignment, and refinement progress.

---

From a FASTA file

from clustalomega._io_6_9 import fasta_oku
from clustalomega import align

sequences = fasta_oku("my_sequences.fasta")
aligned_blocks, names = align(sequences, k=3)

for name, block in zip(names, aligned_blocks):
    print(f"{name:<12} {block}")

---

Command-Line Interface

After installation, run the interactive CLI:

clustalomega

It will ask you to:

1. Choose input method (manual entry or FASTA file)
2. Enter the k-tuple length
3. Run the full pipeline and print all intermediate results

---

Algorithm Overview

The pipeline mirrors the original Clustal Omega algorithm:

1. k-tuple distance matrix
        ↓
2. mBed embedding (reference-based Euclidean projection)
        ↓
3. Bisecting k-means clustering  (⌈√N⌉ clusters)
        ↓
4. UPGMA guide tree
        ├─ per-cluster sub-trees (k-means UPGMA)
        └─ centroid-level super-tree
        ↓
5. Progressive alignment  (Profile HMM + Viterbi)
        ↓
6. Iterative refinement   (HHAlign-style, max 3 rounds)
        ↓
   Final MSA

---

Example: 30-sequence dataset

SP score before refinement : -16487
SP score after  refinement : -12210   (gain: +4277)
Alignment length           : 37 columns

---

Project Structure

clustalomega/
├── __init__.py              # Public API: align()
├── cli.py                   # Interactive command-line entry point
├── _math_utils_1_5.py       # Math helpers (rounding, logarithm, padding)
├── _io_6_9.py               # FASTA parser and manual input
├── _distance_10_15.py       # k-tuple distance matrix
├── _embedding_16_22.py      # mBed embedding
├── _clustering_23_32.py     # Bisecting k-means
├── _guide_tree_33_43.py     # UPGMA guide tree
├── _alignment_44_53.py      # Profile HMM + Viterbi alignment
└── _refinement_54_60.py     # Iterative refinement + SP scoring

---

License

MIT License — see LICENSE for details.

---

Author

Emre Taşdemir — [e-mail]