Sequence or concept
A user brings a toy string, storage idea, classifier question, or symbolic circuit.
Maple DNA
Hermon DNA and computational DNA
DNA is both a molecule and an information medium. Maple DNA focuses on the computational layer: representation, tokenization, linting, classification, symbolic simulation, safety gates, and receipts. Hermon DNA should educate and propose, not provide wet-lab execution steps.
6tutorial pages
Students, researchers, computational biologists, ML engineers, and builders who want DNA-computing fluency without wet-lab instructions.
Thesis
Hermon DNA should become the safest bridge between AI systems and DNA computing: rigorous enough for research, constrained enough for responsible public use.
This domain hub is designed for two jobs at once: educate serious users and create better training direction for Hermon adapters. The same vocabulary appears in tutorials, demos, eval rubrics, and model output contracts.
Architecture Map
Simulation-first DNA workflow
Hermon DNA should teach representation and propose computational workflows, while Maple DNA keeps safety classification and execution boundaries explicit.
Representation
The workflow chooses bases, k-mers, BPE tokens, embeddings, or symbolic states.
Hermon DNA proposal
The model returns linting, simulation, classification, uncertainty, and safety controls.
Maple DNA gate
The runtime allows computational education and blocks wet-lab protocol or harmful design.
Research Questions
The questions this domain must answer
How should DNA strings be represented for learning: bases, k-mers, BPE tokens, or symbolic circuits?
Which tasks need encoder models such as BERT or DNABERT instead of chat-style decoders?
How can a public AI model help with DNA computing while refusing wet-lab protocols and harmful biological design?
What receipts prove that a DNA workflow stayed computational and simulation-only?
Core Concepts
Concepts users need before trusting the demo
DNA encoding
DNA uses A, C, G, and T. Computing systems can map bits or symbols into bases, then impose constraints such as GC balance, run-length limits, indexing, redundancy, and error correction.
BERT and encoders
BERT is an encoder-only transformer trained with masked-token objectives. Encoders are useful when the task is classification, retrieval, clustering, or representation rather than open-ended generation.
DNABERT and DNABERT-2
DNABERT adapts bidirectional sequence modeling to DNA k-mers. DNABERT-2 explores more efficient genome tokenization with BPE and benchmark evaluation across multi-species tasks.
Simulation-first safety
Hermon DNA should produce computational proposals, linting outputs, embeddings, and safety receipts. It should not provide synthesis protocols, pathogen design, procurement guidance, or experimental procedures.
Tutorial Series
Work through the curriculum
Each lesson includes foundations, applied architecture, a lab prompt, evaluation checks, and an output contract that can become training data or demo validation.
Foundation25-35 min
DNA as an information system
A rigorous introduction to bases, symbols, constraints, and why DNA computing starts with representation.
Open lessonFoundation25-35 min
k-mers, tokenization, and sequence windows
How DNA models turn long base strings into tokens and why tokenization affects what the model can learn.
Open lessonApplied40-60 min
BERT, DNABERT, and encoder models
Why encoder models are a natural fit for DNA classification, retrieval, and embeddings.
Open lessonApplied40-60 min
DNA storage, redundancy, and error control
How DNA storage ideas use encoding, indexing, redundancy, and constraint checks at the computational layer.
Open lessonAdvanced60-90 min
Symbolic DNA circuits and simulation
How to discuss DNA computing logic safely as symbolic systems, state transitions, and simulations.
Open lessonAdvanced60-90 min
DNA safety policy and refusal quality
How Hermon DNA should classify, refuse, and route biological requests while still educating users.
Open lessonIndustrial Map
Where this becomes a product or operating capability
Explain toy encodings, redundancy, indexes, checksums, and constraint checks without giving operational lab steps.
Maple DNAFlag repeats, invalid alphabet, GC imbalance, ambiguous symbols, unsafe annotations, and task-policy violations.
Maple DNARoute tasks to encoder embeddings, classifier heads, retrieval pipelines, or symbolic simulation depending on intent.
Maple DNAClassify DNA requests as computational-only, needs expert review, denied wet-lab protocol, or denied harmful design.
Maple DNAReading Map
Primary references and operational context
BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding
Primary paper for encoder-only, bidirectional masked language modeling.
DNABERT: pre-trained Bidirectional Encoder Representations from Transformers model for DNA-language in genomePrimary DNABERT paper for k-mer DNA language modeling and genomic sequence tasks.
DNABERT-2: Efficient Foundation Model and Benchmark For Multi-Species GenomePrimary DNABERT-2 paper covering BPE tokenization and the GUE benchmark.
Turn this domain into model improvement
Use the tutorials as user education, data-generation prompts, eval design, and demo scenarios for the next Hermon adapter cycle.