PathMap, released by Joshua Dungan on July 1, 2026, is a web browser-based, local-first AI bioinformatics engine and a personal literature-based knowledge discovery platform. This tool was developed to facilitate medical literature research, specifically to extract new insights from vast research papers on intractable rare diseases such as Amyotrophic Lateral Sclerosis (ALS). Addressing the inherent limitations of large language models (LLMs), namely information distortion, PathMap boasts an architecture that operates entirely on the client-side. Users can autonomously design knowledge maps tailored to their research objectives within their web browser, without worrying about data leakage or server costs. Existing literature analysis solutions necessarily relied on external cloud APIs, which posed risks of confidential pharmaceutical/medical asset leakage. They also suffered from semantic drift, where LLMs became mired in the probabilistic likelihood between words, creating spurious research facts. This is akin to embarking on a journey without a compass, getting lost in the fog and experiencing hallucinations. To control this, PathMap incorporates a unique verification mechanism called "Veridical Enforcement Measures," which defines the inference limits of generative models. It anchors the AI agent's reasoning path on a verifiable fact trajectory, controlling probabilistic word generation and ensuring that only the rigorous logical structure explicitly stated in the source literature is tracked. In particular, this engine implements Swanson's Puzzle theory, a classic methodology for extracting connections between two or more seemingly unrelated pieces of academic information to discover new hypotheses. Users can explore complex connections between genes, chemicals, and specific diseases, visualize cross-linking genes or interacting pathways, and conduct research such as drug repurposing in a one-stop manner. It also supports easy integration with an Ollama endpoint activated on a local host (localhost:11434), allowing users to seamlessly integrate various open-source LLMs, such as Llama 3 or Phi 3, and process large amounts of information that require overnight processing, all completely free of charge, even in an offline environment. This platform provides an intuitive user environment that supports real-time multilingual translation, significantly reducing the barriers to accessing the latest global research literature on rare diseases. It offers a practical innovation that allows not only large-scale pharmaceutical research labs but also individual researchers focused on specific diseases, and even patient caregivers and independent caregivers, to build the best knowledge network and safely operate RAG pipelines with just one browser.

PathMap, released by Joshua Dungan on July 1, 2026, is a web browser-based, local-first AI bioinformatics engine and a personal literature-based knowledge discovery platform.

🔬 Explore Top ALS Genes Based on TDP-43 Protein Pathway: Using the PathMap web browser in a local environment, connect to Ollama (Llama 3 8B) to analyze approximately 50 ALS publications. Within 30 seconds, map the hidden correlations between genes that induce TDP-43 aggregation and localization, and quantitatively derive candidate top regulatory factors.