Why 22: The Key That Reads the Genome

What is a 22-mer?

A 22-mer is a sequence of 22 consecutive DNA letters. Nothing more. The genome is a text written in a 4-letter alphabet (A, C, G, T), and a 22-mer is a "word" of length 22 in that text.

To compare two genomes, we do something extremely simple:

1. Cut the first genome into every possible 22-letter word (sliding one letter at a time)
2. Store all these words in a table
3. Slide across the second genome, checking each 22-letter word: "Have I seen this before?"
4. Paint every position where a match is found

The result: what percentage of the second genome is covered by words from the first. That's it. No alignment. No gene annotation. No assumptions about what's "important" and what's "junk." The key reads everything equally.

Why 22? The Mathematics of Uniqueness

With 4 letters and 22 positions, there are 4²² = 17,592,186,044,416 possible 22-mers. That's 17.6 trillion possible keys.

A human genome is approximately 3.1 billion base pairs. So the number of actual 22-mers in a genome (~3.1 billion) is roughly 1/5,700th of the total possible space.

This ratio is critical. It means:

• The chance of a random 22-mer matching by accident is approximately 1 in 5,700 (3.1B / 17.6T)
• Each match carries real information — it is not noise
• Yet it is not so stringent that it misses everything

This is the sweet spot. A 20-mer would have 1/90 false hit rate — too noisy. A 30-mer would require near-perfect conservation — too strict, missing all regions with even one mutation. 22 is the minimum key length that produces clean signal from a 3-billion-letter genome.

What Happens With Errors?

DNA accumulates mutations over time. What happens when a single letter changes in a 22-mer?

If position 10 (middle) mutates: the 22-mer at that position breaks — no match. But the 22-mers starting at positions 1-9 (before the mutation) and 12-22 (after) are still intact. A single-point mutation destroys exactly one 22-mer out of the 22 overlapping windows at that site.

Result: 21 out of 22 windows still match = 95.5% survival per single mutation.

If two mutations occur within 22 bases of each other:

• If they're adjacent: most windows still survive (~90%)
• If they're at opposite ends: only windows between them break
• On average: ~86% survival for two mutations in 22 bases

Two errors in 22 bases is a 9% mutation rate — far higher than what's seen between closely related species. So the method is robust against realistic mutation levels, but correctly penalizes heavily diverged sequences.

This is exactly what we want: a key that catches similarity despite noise, but correctly reports divergence when the sequences are truly different.

The Complex Key (Canonical 22-mers)

DNA is double-stranded. Every sequence has a reverse complement on the opposite strand. The sequence ACGTACGTACGTACGTACGTAC on one strand reads GTACGTACGTACGTACGTACGT on the other.

Our tool stores the canonical form of each 22-mer: whichever is alphabetically smaller between the forward sequence and its reverse complement. This means:

• A match on either strand counts
• The effective key space is halved (~8.8 trillion), but still vastly larger than any genome
• False match rate remains below 1 in 2,800 — negligible

This is not a choice. DNA without reverse complement comparison misses half the biology. It would be like reading a book and ignoring every other page.

Why Not Genes? Why Everything?

Traditional genome comparison aligns genes — the protein-coding regions that make up approximately 1.5% of the genome. The other 98.5% is dismissed as "non-coding" or, historically, "junk DNA."

The 22-mer method makes no such judgment. It reads every base equally. This matters because:

1. We don't know what's important. Until 2012, 98.5% of the genome was considered functionless. The ENCODE project showed most of it is biochemically active. Scientists who ignored it were wrong for 50 years.

1. Transposable elements live in the "junk." BovB — the snake-derived element that defines ruminants — is in the non-coding regions. Gene-only comparison would miss it entirely.

1. Regulatory sequences are not genes. The enhancers, promoters, insulators, and piRNA clusters that control gene expression sit in non-coding DNA. They are the software that runs the hardware.

1. 22-mer coverage measures the actual machine, not the manual. Two organisms can have identical genes but completely different regulatory landscapes. The 22-mer sees this. Gene comparison does not.

Consider the Torah analogy: if you compared only the nouns in two manuscripts, you might conclude they are 90% identical. But the grammar — the inflection letters, the prefixes and suffixes, the regulatory structure — is what determines meaning. The 22-mer is the first tool that reads the grammar of the genome, not just the nouns.

The Torah Test

Take the Torah (304,805 letters). Cut it into overlapping 22-letter windows. Scramble the text into random fragments. Run the 22-mer comparison between the original and the scrambled version.

The algorithm will reconstruct nearly all of the original text. Why? Because each 22-letter sequence is a fingerprint. Even scattered across random fragments, the fingerprint persists. The key finds its lock regardless of where the lock was thrown.

This is what we do with genomes. Millions of years of mutation, deletion, insertion, and rearrangement have scrambled the original text. But the 22-mer fingerprints persist wherever the underlying sequence is conserved — and they disappear wherever it has truly diverged.

Every 22-mer match is a verse found intact. Every gap is a verse that was rewritten. The coverage percentage tells you: how much of this text is original, and how much was changed?

The Numbers

The 22-mer sees what gene comparison cannot: that a bee and a wasp, despite looking identical, are 89% different books. And that a pig and a human, despite looking nothing alike, run on the same operating system (L1 = 17.97% vs 17.96%).

No God Required, No Atheism Required

The 22-mer method is agnostic. It does not know what a gene is. It does not know what a transposon is. It does not care about taxonomy, evolution, creation, or design. It counts matches. That's all.

The results speak for themselves. If they align with an ancient text — that is a fact to be examined, not a conclusion to be forced. If they don't — that too is a fact.

We are not God. We do not decide what is junk and what is treasure. We have 22 letters that read everything, and we report what they find. The reader can draw their own conclusions.

Twenty-two letters in the Hebrew alphabet. Twenty-two bases in the key that reads the genome. Both read everything. Both judge nothing.