ABPOA(1)

✨ ABPOA(1) — User Commands ✨

Okay bestie, let's talk about abPOA 💖 (the manual name is abpoa, and it also shows up as abpoa.avx2, abpoa.avx, abpoa.sse4.1, abpoa.ssse3, abpoa.sse3, and abpoa.generic 🦋).

🌸 NAME

abpoa, abpoa.avx2, abpoa.avx, abpoa.sse4.1, abpoa.ssse3, abpoa.sse3, abpoa.generic — adaptive banded Partial Order Alignment ✨

🎀 SYNOPSIS

abpoa [options] <in.fa/fq> > cons.fa/msa.fa/abpoa.gfa

💅 DESCRIPTION

So abPOA is an extended version of Partial Order Alignment (POA) that does adaptive banded dynamic programming (DP) with an SIMD implementation 🔥. It can run multiple sequence alignment (MSA) on a set of input sequences, then make a consensus sequence by applying the heaviest bundling algorithm to the final alignment graph 👑.

Lowkey iconic: it can make high-quality consensus sequences from error-prone long reads and is way faster than existing tools 🚀💜.

It supports three alignment modes (global, local, extension) and flexible scoring schemes that allow linear, affine and convex gap penalties 🧚. It currently supports SSE2/SSE4.1/AVX2 vectorization 💖.

---

🦋 OPTIONS

✨ Alignment ✨

• -m, --aln-mode — alignment mode, an INT, default 0 🌈. 0 is global, 1 is local, 2 is extension.
• -M, --match — match score, an INT, default 2 💖.
• -X, --mismatch — mismatch penalty, an INT, default 4 🙈.
• -t, --matrix — scoring matrix file, a FILE, default Null 🍓. -M and -X are not used when -t is used. Examples: HOXD70.mtx, BLOSUM62.mtx.
• -O, --gap-open — gap opening penalty (O1,O2), an INT(,INT), default 4,24 🔥.
• -E, --gap-ext — gap extension penalty (E1,E2), an INT(,INT), default 2,1 💅. abPOA has three gap penalty modes, where the cost of a g-long gap is:
  • convex (default): min{O1+g*E1, O2+g*E2}
  • affine (set O2 as 0): O1+g*E1
  • linear (set O1 as 0): g*E1
• -s, --amb-strand — ambiguous strand mode, default False 🧚. For each input sequence, it tries the reverse complement if the current alignment score is too low, then picks whichever strand has the higher score 🦋.

✨ Adaptive banded DP ✨

• -b, --extra-b — first adaptive banding parameter, an INT, default 10 🌸. Set b to a value < 0 to disable adaptive banded DP 🙅‍♀️.
• -f, --extra-f — second adaptive banding parameter, a FLOAT, default 0.01 🫶. The number of extra bases added on both sides of the band is b+f*L, where L is the length of the aligned sequence.

✨ Heuristics for graph alignment (under development) ✨

• -G, --int-path-score — include log-scaled path score for graph alignment, default False 🌈.
• -L, --sort-by-len — sort input sequences by length in descending order, default False 💜.
• -R, --gap-on-right — put indels on the right-most side of the alignment (the default is left), default False 🦋.
• -J, --gap-at-end — always put indels at the end of the alignment if possible, default False ✨.

✨ Minimizer-based seeding and partition (only effective in global alignment mode) ✨

• -S, --seeding — enable minimizer-based seeding and anchoring, default False 🎀.
• -k, --k-mer — minimizer k-mer size, an INT, default 19 🧚.
• -w, --window — minimizer window size, an INT, default 10 🌸.
• -n, --min-poa-win — min. size of window to perform POA, an INT, default 500 💖.
• -p, --progressive — build a guide tree and perform progressive partial order alignment, default False 👑.

✨ Input/Output ✨

• -Q, --use-qual-weight — take the base quality score from a FASTQ input file as the graph edge weight for consensus calling, default False 🔥. This is only effective when input sequences are in FASTQ format and consensus calling uses heaviest bundling.
• -c, --amino-acid — input sequences are amino acid (the default is nucleotide), default False 🧬.
• -l, --in-list — input file is a list of sequence file names, default False 📋. Each line is one sequence file containing a set of sequences, which abPOA will align to generate a consensus sequence 🫶.
• -i, --incrmnt — incrementally align sequences to an existing graph or MSA, a FILE, default Null 🦋. The graph can be in GFA or MSA format generated by abPOA.
• -o, --output — output to FILE, default stdout 💌.
• -r, --result — output result mode, an INT, default 0 🌈. The modes are:
  • 0: consensus in FASTA format
  • 1: MSA in PIR format
  • 2: both 0 & 1
  • 3: graph in GFA format
  • 4: graph with consensus path in GFA format
  • 5: consensus in FASTQ format
• -a, --cons-algrm — consensus algorithm, an INT, default 0 💅. The options are:
  • 0: heaviest bundling path in partial order graph
  • 1: most frequent bases at each position
• -d, --maxnum-cons — max. number of consensus sequences to generate, an INT, default 1 👯‍♀️.
• -q, --min-freq — min. frequency of each consensus sequence (only effective when -d/--num-cons > 1), a FLOAT, default 0.25 🌷.
• -g, --out-pog — dump final alignment graph to FILE (.pdf/.png), default Null 📸.
• -h, --help — print this help usage information 🆘.
• -v, --version — show version number 🔢.
• -V, --verbose — verbose level, an INT, default 0 🗣️. 0 is none, 1 is information, 2 is debug.

🌈 SEE ALSO

For more information, please refer to the paper published in Bioinformatics 📖:

https://dx.doi.org/10.1093/bioinformatics/btaa963

---

Bye bestie, that's the whole tea ☕💖 — abpoa 1.5.5, November 2025, ABPOA(1) 🫶✨