A Typed Programming Language for Biological Systems

DNA is not one type — it contains seven functional data types that compose into a complete programming model. DNA-Lang formalizes each as a typed construct, implements them in Haskell, and synthesizes a unified orchestration language for biological intent, AI prediction, genome editing, and regulatory traceability.

Total: 207 pages across 8 papers

data Genome = Genome
  { code        :: [ProteinCode]      -- Paper I:   Executable Functions
  , regulators  :: [Regulator]        -- Paper II:  Control Flow
  , ncRNAs      :: [RNAControl]       -- Paper III: Signals & Middleware
  , structure   :: [Structure]        -- Paper IV:  Memory Layout
  , repeats     :: [Repeat]           -- Paper V:   Self-Modifying Code
  , epigenetics :: [State]            -- Paper VI:  Runtime State
  , programs    :: [Program]          -- Paper VII: Orchestration
  }

DNA-Lang is a typed orchestration language — not "DNA as code" directly — that spans:

1. Biological intent — declare targets, edits, assays
2. AI model invocation — typed calls to AlphaFold, structure predictors, guide rankers
3. Sequence/edit design — CRISPR guide design, donor templates, vector construction
4. Experimental execution — assay planning, controls, readouts
5. Safety/constraint checking — off-target scoring, payload limits, toxicity thresholds
6. Provenance and regulatory traceability — full decision log, model versions, sequence lineage

Layer 1: IR / Schema          — typed intermediate representation
Layer 2: DSL                  — declarative language (design, edit, screen, validate)
Layer 3: Target Compilers     — SBOL, Benchling, AlphaFold, LIMS/ELN
Layer 4: Verification         — off-target, payload, manufacturability, provenance

Four compiler artifacts:

• Design — sequences, guides, constructs, edit plans
• AI Pipeline — structure predictions, rankings, uncertainty estimates
• Lab Protocol — assay plan, controls, acceptance criteria
• Compliance — provenance, model versions, decision log

cd papers/latex
pdflatex -interaction=nonstopmode <paper>.tex
pdflatex -interaction=nonstopmode <paper>.tex  # second pass for refs

cd src/<subject>
ghc -o main Main.hs
./main

papers/latex/    — 8 LaTeX source files
papers/pdf/      — 8 compiled PDFs
src/             — Haskell implementations (one directory per paper)
docs/            — GitHub Pages site
website/         — Next.js/Vercel site
scripts/         — Build tools (latex2html.py, paper-template.html)

Matthew Long — The YonedaAI Research Collective — Chicago, IL — 2026

• Vercel: https://dna-lang.vercel.app
• GitHub Pages: https://yonedaai.github.io/dna-lang/
• Papers: available as PDF and HTML with KaTeX math rendering