RENKIN — Retrosynthetic Exploration Network for Knowledge-Informed Navigation

Computer-Aided Synthesis Planning (CASP) · Pure Rust · WebAssembly · Python
Named after 錬金 (れんきん, renkin) — Japanese for alchemy: just as alchemists transformed base metals into gold, RENKIN transforms target molecules back into cheap starting materials.

日本語版 README · Documentation · Live Demo →

What is RENKIN?

RENKIN is an open-source retrosynthesis engine for computer-aided synthesis planning (CASP) that automatically discovers optimal chemical reaction routes from a target molecule back to cheap, commercially available starting materials.

Built entirely in Rust with the chematic cheminformatics crate. Zero C/C++ dependencies. All crates enforce #![forbid(unsafe_code)] — compiler-verified Pure Safe Rust throughout.

→ Try the Live Playground — runs entirely in WebAssembly, no installation needed.
→ Full Documentation — API reference, examples, benchmark.

Installation

pip install renkin          # Python
cargo add renkin            # Rust
npm install renkin          # JavaScript / Node.js

Quick Start

import renkin

result = renkin.find_routes(
    "CC(=O)Oc1ccccc1C(=O)O",   # Aspirin
    depth=5,
    max_routes=3,
)

for route in result["routes"]:
    for step in route["steps"]:
        print(f"  {step['target']} → {' + '.join(step['precursors'])}  [{step['rule']}]")

import init, { find_routes } from './pkg/renkin.js';
await init();
const result = JSON.parse(find_routes("CC(=O)Oc1ccccc1C(=O)O", 5, 3, 0));

./target/release/renkin --target "CC(=O)Oc1ccccc1C(=O)O" --depth 5 \
    --templates data/templates_extracted_5000.smi --format tree

Target: CC(=O)Oc1ccccc1C(=O)O
Routes found: 3

Route 1  [score=1.02, depth=1]
OC(=O)c1ccccc1OC(=O)C
└── [extracted_169]
    ├── OC(=O)C  ✓ BB
    └── [OH]c1ccccc1C(=O)O  ✓ BB

Route 2  [score=1.02, depth=1]
OC(=O)c1ccccc1OC(=O)C
└── [extracted_145]
    ├── CC(=O)Cl  ✓ BB
    └── [OH]c1ccccc1C(=O)O  ✓ BB

Route 3  [score=1.03, depth=1]
OC(=O)c1ccccc1OC(=O)C
└── [extracted_238]
    ├── c1cccc(c1O)C(O)=O  ✓ BB
    └── C([OH])(=O)C  ✓ BB

Use --format mermaid for GitHub/Notion-compatible flowcharts.

Constraint-based Search

Restrict routes by the element composition of their building blocks.

Default search — all 5 routes for biphenyl:

renkin --target "c1ccc(-c2ccccc2)cc1" --templates data/templates_extracted_5000.smi --format tree

Routes found: 5
Route 1  [score=1.00, depth=1]  c1ccccc1Br + c1c(B(O)O)cccc1
Route 2  [score=1.03, depth=1]  c1ccccc1Br + c1c(B(O)O)cccc1
Route 3  [score=1.06, depth=1]  c1cc(Cl)ccc1 + c1c(B(O)O)cccc1
Route 4  [score=1.08, depth=1]  c1(I)ccccc1  + c1c(B(O)O)cccc1
Route 5  [score=1.08, depth=1]  c1ccccc1Br  + c1(B2OC(C(C)(C)O2)(C)C)ccccc1

Constrained search — boronic-acid coupling, no Br or I starting materials:

renkin --target "c1ccc(-c2ccccc2)cc1" --templates data/templates_extracted_5000.smi \
    --require-elements "B" --avoid-elements "Br,I" --format tree

Routes found: 1

Route 1  [score=1.06, depth=1]
c1ccccc1-c2ccccc2
└── [extracted_398]
    ├── c1cc(Cl)ccc1  ✓ BB
    └── c1c(B(O)O)cccc1  ✓ BB

Constraints compose freely. Applied as a post-filter on completed routes — the A* search itself is unchanged.

Add --verbose to print search statistics (nodes expanded, elapsed time) to stderr. Performance counters are available in native builds only; disabled in WASM.

Key Features

Benchmark

USPTO-50k test set (4,907 molecules, full evaluation):

Evaluation note: All numbers use the standard USPTO-50k train/test split (same corpus). Templates are extracted from the training set and evaluated on the test set — the same methodology as AiZynthFinder and other published tools. Numbers reflect performance within the USPTO-50k domain; out-of-distribution generalization has not been separately evaluated.

* 29/50 chunks, previous binary
† 50/50 chunks — 72.1% (3,540/4,907) confirmed

On the standard USPTO-50k benchmark (multi-step route-finding, same train/test split), RENKIN (78.1%) exceeds the published numbers for AiZynthFinder (45–53%), Retro* (44.3%), and ASKCOS (41%) — though those are from 2019–2020 papers with different BB/template counts, so no matched-condition experiment exists yet.
Note: LocalRetro (53.4%) and GLG (58.0%) report single-step top-1 prediction accuracy — a different metric, not directly comparable.
Full benchmark details →

Competitive Landscape

RENKIN's goal: match or exceed neural-network-based tools using only curated rules and auto-extracted SMIRKS templates — no GPU, no training data, no black boxes. On the standard USPTO-50k benchmark (same train/test split used by all published tools), RENKIN reaches 78.1% (3,831/4,907 — full 4,907-molecule run confirmed). Template frequency weighting (Phase A) — the same principle as AiZynthFinder's neural template scoring — combined with 5,000 auto-extracted templates and 509 building blocks delivers this result. RENKIN runs anywhere: browser, CLI, Python — single cargo build.

MCP Server

renkin-mcp exposes retrosynthesis as an MCP tool so AI agents (Claude, etc.) can call it directly.

Setup — add to claude_desktop_config.json:

{
  "mcpServers": {
    "renkin": { "command": "/path/to/renkin-mcp" }
  }
}

Tool: find_routes(smiles, depth?, max_routes?, avoid_elements?, require_elements?)

The server auto-detects data/building_blocks.smi and data/templates_extracted_5000.smi in the working directory. Falls back to the embedded 509-BB / 20-rule defaults if not found.

cargo build --release
# binary: target/release/renkin-mcp

Architecture

Target SMILES
     │
     ▼
┌─────────────────────────┐
│     chem_env.rs         │  ← chematic wrapper
│  - SMILES parse         │     canonical-SMILES FxHashSet BB lookup (O(1))
│  - 5,000 retro rules    │     fragment sanitization + ring-leak filter
│  - Building block check │     apply_retro memoization cache
└────────────┬────────────┘
             │  par_iter (rayon / sequential on WASM)
             ▼
┌─────────────────────────┐
│      search.rs          │  ← A* / AND-OR Tree Search
│  - Priority queue       │     SA Score heuristic + memoization
│  - Closed list          │     beam search (SmallVec frontier)
│  - Arc<PathNode> paths  │     O(1) path sharing per child
└────────────┬────────────┘
             │
             ▼
┌─────────────────────────┐
│      score.rs           │  ← Heuristic / Cost Function
│  - SA Score (chematic)  │     h = Σ(1 + 0.5·(sa−1)/9)
│  - MW step cost         │     g = Σ(1 + total_mw/2000)
└────────────┬────────────┘
             │
             ▼
┌─────────────────────────┐   (optional)
│      scorer.rs          │  ← Phase B: NN Template Scorer
│  - tract-onnx           │     Pure Rust ONNX inference
│  - --scorer flag        │     molecule-specific template ranking
└────────────┬────────────┘
             │
             ▼
  JSON  ←  CLI / Python / WASM

Project Structure

renkin/
├── Cargo.toml
├── src/
│   ├── lib.rs               # public library
│   ├── main.rs              # CLI binary  (--templates, --scorer flags)
│   ├── bin/benchmark.rs     # renkin-bench binary  (--templates flag)
│   ├── chem_env.rs          # 5,000 retro rules, BB check, template loader
│   ├── score.rs             # SA Score heuristic + step cost
│   ├── search.rs            # A* / AND-OR tree engine + beam pruning
│   ├── scorer.rs            # Phase B: tract-onnx NN template scorer
│   ├── python.rs            # PyO3 bindings (--features python)
│   └── wasm.rs              # wasm-bindgen bindings (cfg = wasm32)
├── data/
│   ├── building_blocks.smi              # 480 curated commercial starting materials
│   ├── templates_extracted_5000.smi     # 5,000 auto-extracted SMIRKS templates
│   ├── benchmark_targets.smi            # internal benchmark set
│   └── bench_chunks/                    # USPTO-50k per-chunk results
├── scripts/
│   ├── extract_templates.py         # rdchiral template extraction pipeline
│   └── run_benchmark_chunks.sh      # resumable chunked benchmark runner
├── docs/                # MkDocs source → kent-tokyo.github.io/renkin/
└── mkdocs.yml

Roadmap

• Route cost scoring (commercial reagent price integration)

• SMIRKS retro-reaction rules + fragment sanitization
• A* / AND-OR tree search, closed list, degenerate-route filter
• SA Score heuristic + beam search
• Parallel rule application (rayon; sequential fallback on WASM)
• Python bindings (PyO3 + maturin) · pip install renkin
• WASM build · npm install renkin
• Benchmark CLI (renkin-bench) + USPTO-50k evaluation
• WASM browser playground + i18n (EN/JA/ZH)
• Graph-based biaryl cleavage · O(1) canonical-SMILES BB index
• Published to crates.io / PyPI / npm · GitHub Actions CI/CD
• MkDocs documentation site · GitHub Pages playground
• Auto template extraction (rdchiral): 27.8% → 78.1% USPTO-50k
• Tetrahedral stereo @/@@ + E/Z double-bond stereo
• Template frequency weighting (Phase A): 72.1% USPTO-50k
• FxHashMap · SmallVec beam frontier · SA Score memoization · Arc path sharing
• 5,000 extracted templates + 509 BBs: 78.1% USPTO-50k (3,831/4,907 ✅)
• NN template scorer via --scorer flag (tract-onnx, Pure Rust ONNX)
• --format tree|mermaid route visualization
• Constraint-based search: --avoid-elements, --require-elements
• --verbose search statistics to stderr
• MCP server (renkin-mcp) — AI agents call retrosynthesis directly
• #![forbid(unsafe_code)] — compiler-enforced Pure Safe Rust

Citation

If you use RENKIN in academic work, please cite:

@software{renkin2026,
  author    = {kent-tokyo},
  title     = {{RENKIN}: Retrosynthetic Exploration Network for Knowledge-Informed Navigation},
  year      = {2026},
  url       = {https://github.com/kent-tokyo/renkin},
  version   = {0.1.4},
  license   = {MIT}
}

License

MIT

GitHub Topics: retrosynthesis cheminformatics wasm rust drug-discovery casp synthesis-planning computational-chemistry