The Cell — DNA → mRNA → Protein → 3D Fold
Type a DNA sequence. Watch it transcribe to mRNA, translate to amino acids using the actual genetic code, then fold into a 3D protein with hydrogen bonds and hydrophobic core. Real molecular biology.
What this is
The central dogma of molecular biology, executable in your browser. Type a DNA sequence (or pick from 5 real preloaded ones — preproinsulin, hemoglobin β, zinc finger, GFP, ubiquitin). Watch the page transcribe it to mRNA, walk a ribosome along the mRNA emitting amino acids via the 64-codon table, then fold the resulting chain in 3D using Levinthal- style energy minimization. Backbone as ribbon. Hydrogen bonds as cyan dashes. Hydrophobicity colored. Run for 300 iterations and the structure collapses into something protein-shaped.
Why this is mind-blowing
This is the kind of pipeline that took NIH labs decades to build visualizers for. You wrote it from one paragraph. It runs in the browser. No CUDA, no AlphaFold, no install. The fold isn't physically accurate — but it's directionally correct, and the educational value is enormous. Students can SEE why the genetic code is what it is.
Build a single-file biology demo. Input box for a DNA sequence. The
page transcribes it to mRNA (T→U), translates it to amino acids using
the standard genetic code (the actual 64-codon table), then folds the
resulting protein chain into a 3D structure using simplified Levinthal-
style energy minimization (hydrogen bonds, hydrophobic core, backbone
ribbons). Render the folded protein in WebGL with side chains as spheres.
Show the central dogma — DNA → RNA → protein → fold — as a visual
pipeline at the top.Paste this into Claude, Cursor, or Copilot. Change one thing that matters to you.
What I learned shipping it
- The 64-codon genetic code is one of the most-trained pieces of trivia in the model's head. You don't have to provide it. You just have to ask for it correctly.
- Energy minimization for protein folding doesn't need quantum chemistry. Hydrogen bonds (distance-based attraction), hydrophobic burial (pull hydrophobic residues toward chain centroid), and a simple repulsion term get you 80% of the visual story for 1% of the compute.
- Real preloaded sequences (insulin, hemoglobin, GFP, ubiquitin, zinc finger) make the demo educational instead of a toy. The model knows them all by heart.