Products / Language Models EC 2.7.- Phosphotransferase Superfamily

A protein language model, built from scratch — on a laptop.

NaviKinase 1.0 is a 149.6M-parameter decoder-only transformer trained entirely locally on Apple Silicon via MLX — no cloud compute, no fine-tune of an existing model. It speaks the language of kinases: 2.45M UniProtKB sequences, 1.32B tokens, one enzyme superfamily.

navikinase — generate.py
149.6M
PARAMETERS
2.45M
TRAINING SEQUENCES
1.32B
TRAINING TOKENS
4.32
HELD-OUT TEST PERPLEXITY
01

Architecture & evaluation

NaviKinase 1.0 model specifications
ArchitectureDecoder-only transformer
NormalisationPre-norm RMSNorm
Position encodingRoPE
Layers18
d_model832
Attention heads13 (head dim 64)
FFN dimension3,328
Context length1,024 tokens
Precisionfp32
FrameworkMLX

Held-out, not cherry-picked

The test set was never touched during training or checkpoint selection — validation handled that, which is why it carries a small optimistic bias. The test number is the one to cite.

Best validation perplexity3.98
Held-out test perplexity4.32
Held-out test loss1.4633

Trained the full 40,289-step ceiling (2 epochs) to completion — no early stopping triggered. Val loss kept improving at every eval interval.

02

What it can do

NaviKinase doesn't speak English — there's no chat template, no instruction-following. It speaks amino acid sequences. Everything around it is tooling that makes that useful.

Sequence generation

Generate from scratch, condition on one of 14 EC 2.7.- sub-subclasses, or continue a real sequence you provide.

Plausibility scoring

Teacher-forced loss over a real sequence — a plausibility signal for how "kinase-like" it reads to the model.

Plain-English front end

A pattern-matching router (not an LLM) that maps natural requests onto the tools — and asks rather than guesses when it matters.

Live lookups

Real UniProt accession fetches and PubMed citation search — the only parts of this release that need internet access.

Research pipeline

"Research protein X" chains fetch → score → classify → motif scan → literature search into one synthesised report.

Motif scanning

Regex checks for canonical catalytic motifs — P-loop, HRD, DFG — as a cheap directional signal, not a benchmark.

3D structure

Real structure prediction via ESMFold — a genuinely different model, called on NaviKinase's output or any fetched sequence.

Non-synthesis notice

Every fresh generation carries an automatic notice: treat output as an unvalidated hypothesis, not a confirmed protein.

03

How it was built

No cluster, no cloud bill. NaviKinase was trained unattended over several days on a MacBook.

  • 3.1

    Data at the cluster level

    UniProtKB entries with EC 2.7.-, 30–2000 residues, non-fragment — deduplicated at the UniRef90 cluster level so near-identical bacterial strain variants can't dominate training, with train/val/test split kept at the same cluster level.

  • 3.2

    EC-tag conditioning

    Each example is packed as <BOS> <EC-tag> residues <EOS>, conditioning generation on one of 14 EC 2.7.- sub-subclasses, or <EC_UNKNOWN> for the ~3.3% without a recognised one.

  • 3.3

    AdamW, cosine schedule

    Betas 0.9/0.95, weight decay 0.1, peak LR 3e-4, effective batch size 64 (batch 4 × grad-accum 16) over a 1,024-token context.

  • 3.4

    Built to survive a laptop

    Automatic checkpointing, power-aware pause/resume, and crash recovery via launchd — designed for unattended runs on constrained hardware without a dedicated server environment.

04

Scope, limits & responsible use

Scope is EC 2.7.- only: no exposure to other transferase classes, non-enzymatic proteins, or non-protein biomolecules.
Character-level, single-sequence model — no structural, MSA, or homology information, only raw sequence statistics.
Motif-presence checks are a cheap directional signal, not a validated structural or functional benchmark.
chat.py is keyword/regex pattern matching, not a language model, and says so rather than guessing on an unrecognised request.
score.py --classify is an improvised proxy — the model was never trained as a classifier.
UniProt name search is best-effort free-text ranking, not guaranteed to surface the intended protein first.
Dual-use disclosure — this repo has no content moderation, toxin/hazard blocklist, or biosecurity screening, and has not been reviewed by a biosecurity expert. The realistic risk is limited by scope — EC 2.7.- kinases are among the most common, well-characterised enzyme families, present in nearly all organisms and not associated with toxicity — but that narrowness came from data and model-size decisions, not an engineered safety feature. Before any wet-lab use: route synthesis through a provider that performs standard biosecurity screening, and independently review function and safety yourself.
05

Licence

CC BY-NC 4.0

Free for non-commercial use

Use, share, and adapt NaviKinase 1.0 freely for non-commercial purposes with attribution. Weights, code, and tokeniser are covered.

Commercial licensing

Building a product on this?

Any use intended to generate revenue requires prior written permission. Reach out and we'll work out terms.

jaydev.bhatt@navitratech.com

Trained for a laptop. Ready for yours.

Pull the weights, run it locally with MLX, and see what a purpose-built, single-family protein model can do.