njswalig

Needleman–Wunsch (global) and Smith–Waterman (local) pairwise sequence alignment in plain Python, with no dependencies. It prints the alignment and, if you ask, the filled-in dynamic-programming score table.

pip install njswalig

Usage

from njswalig import njalign      # Needleman-Wunsch (global)
from njswalig import swalign      # Smith-Waterman (local)

seqs = ["ATCG", "TTTA"]           # any number of sequences, any lengths
scores = [2, -1, -2]              # [match, mismatch, gap] -- always this order

alignment = njalign.sequences(seqs, scores)   # or njalign.sequence(...) -- same thing
alignment.show()

Needleman-Wunsch (global): Seq1 vs Seq2, score = -1
Seq1  ATCG
      .|..
Seq2  TTTA

| marks a match, . a mismatch and a space a gap.

Smith–Waterman

swalign.sequences(["TGTTACGG", "GGTTGACTA"], [3, -3, -2]).show()

Smith-Waterman (local): Seq1 vs Seq2, score = 13
Seq1  2 GTT-AC 6
        ||| ||
Seq2  2 GTTGAC 7

The numbers show where the aligned region starts and ends in each sequence (1-based).

More than two sequences

Every pair is aligned. Sequences are named Seq1, Seq2, … in list order, or pass your own:

njalign.sequences(["GATTACA", "GCATGCU", "GATCA"], [1, -1, -1], names=["a", "b", "c"]).show()

The DP score table

swalign.sequences(["ATCG", "TTTA"], [2, -1, -2]).show(matrix=True)

  - T T T A
- 0 0 0 0 0
A 0 0 0 0 2
T 0 2 2 2 0
C 0 0 1 1 1
G 0 0 0 0 0

Using the results in code

sequences(...) returns a list of pair alignments. Each has score, aligned1, aligned2, match_line, name1, name2, seq1, seq2, matrix (the DP table as a list of lists) and 1-based start1, end1, start2, end2.

best = njalign.sequences(["GATTACA", "GCATGCU"], [1, -1, -1])[0]
print(best.score, best.aligned1, best.aligned2)     # 0 G-ATTACA GCA-TGCU

Details

• Scoring is linear: every gap position costs gap. scores must be a list of exactly three numbers, [match, mismatch, gap]; anything else raises ValueError.
• Sequences are stripped of surrounding spaces and upper-cased, so "acgt " equals "ACGT".
• Ties: when several alignments share the best score, one is returned. The traceback prefers a diagonal step (match/mismatch), then a gap in the second sequence, then a gap in the first. Smith–Waterman starts from the first highest cell, scanning row by row.
• No local match: when nothing scores above 0, Smith–Waterman reports an empty alignment with score 0.

License

MIT