# Retrosynthesis {% hint style="info" %} Pending AI’s various capabilities are powered by artificial intelligence which is an experimental technology and may occasionally be misleading or incorrect. {% endhint %} Pending AI's Retrosynthesis Engine is a cutting-edge service designed for ultra-high-throughput synthetic feasibility analysis. By integrating multiple unique deep learning models with a Monte Carlo Tree Search (MCTS) framework, the Engine rapidly identifies complex, multi-step, and economically viable synthetic routes for novel and existing chemical entities. The synthetic bottleneck is a major impediment in drug discovery. Our engine drastically accelerates the Design-Make-Test-Analyze (DMTA) cycle by providing chemist-first, feasible synthesis routes early in the process. * **Intelligent Feasibility**: Combines reaction pattern recognition from deep learning models with the exhaustive search capabilities of MCTS to propose routes that are both practically possible and literature-inspired. * **Massive Throughput**: Designed to scale and screen the synthetic accessibility of hundreds of thousands of compounds per day, enabling proactive filtering of synthetically difficult molecules *prior* to the Make stage. * **Actionable Outputs**: Provides detailed, step-by-step route trees, including learned scoring and pairwise diversity to guide lab-based synthesis. See the **PAI Retro** page for more information about the service. {% content-ref url="../api-reference/pai-retro" %} [pai-retro](https://docs.pending.ai/api-reference/pai-retro) {% endcontent-ref %} *** ## Applications High-throughput large-scale molecule retrosynthesis has traditionally blocked drug discovery pipelines forcing chemists to perform insufficient workarounds for a target. The [PAI Retro](https://docs.pending.ai/api-reference/pai-retro) service provides an avenue to unlock previously time-restricted applications within the DMTA cycle:


Assess Synthetic Accessibility	Proactively filter lead candidates from virtual screening campaigns (or sampled synthetic libraries using generative-ai).
Route Design & Optimisation	Identify cheaper, shorter, or more scalable routes for existing commercial products from specified building block libraries.
Targeted Proprietary Libraries	Retrosynthesis models can be customised using proprietary datasets for reaction templates or building block libraries.
Automated Synthesis Integration	Directly feed valid routes into automated synthesis platforms, electronic laboratory notebooks, or AI-driven environment.

*** ## Features & Advantages ### 1. Chemically-Aware AI Architecture Our unique approach integrates three distinct, high-quality AI models trained on chemically diverse datasets within an engine-variant MCTS search framework. * **Reaction Prediction AI**: Base models are trained on millions of unique chemical reactions to accurately predict viable precursors in a single retrosynthetic step. * **Template Extraction**: Over 75,000 high-quality, chemically relevant reaction templates are extracted and utilised, ensuring outputs reflect real-world reaction space. * **AI-Driven Expansion Policy**: A highly optimised AI model (trained on 71 million samples) guides MCTS expansion, prioritising ideal reactions for a given target molecule for faster convergence. ### 2. Scalability and Customisation We offer flexible solutions designed to meet both the scale of virtual screening and the specificity of proprietary data. * **Ultra-High Throughput**: Standard processing capacity of up to 400,000 queries per day, with tiered options (High, Very-High, Ultra-High) available for bespoke campaigns. * **Custom Data Integration**: The AI models can be customised and fine-tuned using proprietary datasets (e.g., in-house electronic laboratory notebook data), ensuring generated routes prioritise preferred, reliable, and available chemistry. * **Building Block Awareness**: Routes are assessed against a comprehensive set of nearly ten million commercially available building blocks (including a subset of ZINC labelled as in-stock). ### 3. Chemist-First Capabilities The Retrosynthesis engine is designed to augment, not replace, the medicinal chemist's expertise by providing practical, high-value insights. * **Route Diversity Analysis**: MCTS is engineered to explore a rich set of distinct synthetic pathways. Diversity rankings and analysis of routes allows chemists to assess trade-offs in terms of cost, novelty, and other characteristics. * **Feasibility Scoring**: Routes and reaction steps are assigned confidence scores allowing the chemist to quickly prioritise higher-confidence pathways and assess risks for lower scoring components. --- # Agent Instructions: Querying This Documentation If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question. Perform an HTTP GET request on the current page URL with the `ask` query parameter: ``` GET https://docs.pending.ai/capabilities/retrosynthesis.md?ask= ``` The question should be specific, self-contained, and written in natural language. The response will contain a direct answer to the question and relevant excerpts and sources from the documentation. Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.