Gap Reliability Panel

GapCheck Gap-aware conservation diagnostics for alignment columns
Karagöl and Karagöl 2026 (version 1.0-web)

Alignment input nothing leaves this page

An entropy conservation score that counts the gap as a twenty-first symbol rises with gappiness: a column that is mostly gap looks conserved because the gap itself is the frequent symbol. Paste an aligned FASTA to see each column's gap fraction, how many comparators are actually left, and what the score becomes when it is recomputed over the residues alone.

Every sequence must be the same length, which is what makes it an alignment. The file is read and the arithmetic is done in this page, so no sequence is uploaded anywhere. To build the alignment first, BioAlign does it in the browser from the same toolkit.

Across the alignment

Mean conservation by gap decile

as reported, gap as a symbol recomputed over residues only
Mean conservation against the gap fraction of the column, as reported and recomputed over residues only

Columns

Method

The two definitions

Conservation as reported

1 - H / log2(21), H over the twenty residues and the gap

The gap is the twenty-first symbol. A column that is mostly gap has low entropy over that alphabet, so it scores as conserved.


Recomputed over residues only

1 - H / log2(20), H over the residue frequencies renormalised to the sequences that carry a residue

The gap is dropped and the remaining residues are renormalised, so the score describes the comparators that exist.

The two flags

A column is called gap inflated above a gap fraction of 0.25, which is the point past which the column is not reliably alignable, and depth starved below five effective comparators, where there is no conservation estimate to make. The gap fraction is the share of sequences carrying a gap; the effective comparator count is the number carrying a residue.

What it will not match exactly

Frequencies here are unweighted, one count per sequence, over the columns of the alignment as pasted. A pipeline that weights sequences by redundancy, trims columns by coverage, or indexes positions against a target sequence instead of the alignment will give different values from the same sequences. Symbols outside the twenty residues and the gap, such as X or B, are counted separately and left out of both the gap fraction and the entropy.

Method and code

Karagöl and Karagöl, two opposite failure modes in alignment-based conservation scores, one inflating and one concealing constraint, are computable before the score is used. The analysis code is at github.com/karagol-taner/conservation-score-reliability. This page is released under the MIT license, copyright 2026 Taner Karagöl and Alper Karagöl.