The Map Is You: a Bio Beats song

A 2 minute 16 second hip-hop song about the maps and atlases of human biology. Animated diagrams play behind karaoke lyrics.

How it teaches (Justin Sung style)

The song follows Justin Sung's "big picture first, then layers, then links" approach, the same idea behind ../al-qalam-sung-memorization/:

Big picture first. The intro sets up one frame before any details: the central dogma, DNA → RNA → Protein, plus what lies beyond it. A row of four chips at the top of the screen stays visible for the whole video and lights up the layer each scene is about.
One layer per verse. Verse 1 covers the DNA maps. Verse 2 covers the switches and RNA maps. Verse 3 covers proteins and the maps beyond the dogma.
Revisit the whole picture. Each chorus shows the layer stack again and fills in only the maps learned so far. Layers not yet covered show "?".
Link, don't list. The outro is a connected map. Each atlas leads to the next through the question it leaves open, for example "gaps?" from the Genome Project to T2T, and "RNA ≠ protein" from HuBMAP to the Protein Atlas. These links are a teaching device, not the historical order. The Protein Atlas, for example, began in 2003, before the Cell Atlas.

Part	Topic
Intro	Big picture first: every map of you is a layer on one idea.
Verse 1 · DNA	The central dogma: Crick, 1958, DNA → RNA → protein, one way. The Human Genome Project (1990–2003) read about 3 billion letters but left about 8% unfinished. Telomere-to-Telomere filled the gaps in 2022. The pangenome added more people: 47 in 2023 and 232 in 2025. Only about 1–2% of the genome codes for protein, across about 20,000 genes.
Chorus	DNA, RNA, protein. Each map is a layer, and stacked together the layers are you.
Verse 2 · Switches & RNA	Cells share one genome but use different genes. ENCODE's registry lists 2.37 million candidate switches (promoters and enhancers). GTEx measured RNA in 54 tissue sites from 948 donors. For example, albumin is very active in the liver and nearly silent in the brain. The Human Cell Atlas (since 2016) reads RNA cell by cell. HuBMAP (NIH, 2018) records where each cell sits in a tissue.
Verse 3 · Proteins & beyond	The Human Protein Atlas (Sweden, since 2003) uses antibodies to show where proteins sit inside cells. The Human Microbiome Project (NIH, 2007) mapped the body's microbes. Bacteria and human cells are about 1:1 (not 10:1), but gut microbe genes outnumber ours about 150×. The connectome: 1 mm³ of human cortex (2024) holds about 57,000 cells and 150 million synapses, and the data takes up 1.4 petabytes. A whole brain is over a million times bigger.
Outro	Big picture first, then layer on the view, then link every map.

ENCODE4 registry, 2.37 million human cCREs covering about 21% of the genome: Moore et al., Nature, 2026.
Pangenome: 47 people (Nature, 2023), and Release 2 with 232 people (May 2025).
Human Cell Atlas, founded 2016, over 100 million cells from over 10,000 people: HCA press release, Nov 2024. The ~37 trillion cells figure is the estimate the HCA itself quotes.
GTEx V8, 54 tissue sites and 948 donors: NIH Common Fund.
HuBMAP, first awards in 2018: NIH.
Human Protein Atlas, started 2003 in Sweden, antibody-based: proteinatlas.org/about.
Human Microbiome Project, launched December 2007: NHGRI.
Bacteria and human cells: about 38 trillion bacteria and 30 trillion human cells, a ratio of about 1.3:1. Sender, Fuchs & Milo, PLOS Biology, 2016.
1 mm³ of human cortex: about 57,000 cells, about 150 million synapses, and 1.4 PB of data. Harvard describes the sample as about half a rice grain. Shapson-Coe et al., Science, 2024, Harvard Gazette.

These are standard textbook facts, used without a separate source check:

Crick stated the central dogma in 1958.
The Human Genome Project ran from 1990 to 2003.
About 8% of the genome was still missing after 2003, which T2T fixed in 2022.
About 1–2% of the genome codes for protein, across about 20,000 protein-coding genes.
Albumin is made almost only in the liver.
The gut microbe gene catalog is about 3.3 million genes (MetaHIT, 2010).
A brain is about 1.2 million mm³.

This uses the same pipeline as ../polyphenols-song/:

lyrics.json: the lyrics, the bar each line starts on, the tempo (84 BPM), the chord loop (F–G–Em–Am), and the four chorus melodies. A line can have a say field, which is the spelling sent to text-to-speech. For example, T2T is sent as T-two-T.
build_audio.py: makes the voice with Microsoft Edge TTS (en-US-AndrewMultilingualNeural) and places each line on the bar grid. Chorus lines are re-pitched to the melody with the WORLD vocoder (pyworld). The beat is synthesized with numpy. It writes song.wav, timeline.json (start and end times for every lyric word), and song.srt.
scene.html: draws every frame on a canvas from render(t) and timeline.json.
render.mjs: renders 24 fps frames with headless Chrome and pipes them to ffmpeg.
build.sh: runs all of the above.

CHROME=/path/to/chrome src/build.sh

It needs uv, ffmpeg, Node, and network access for Edge TTS and Google Fonts. Intermediate files go to $OUT (default /tmp/maps).