Spaces:
Running
Running
| <html lang="en"> | |
| <head> | |
| <meta charset="UTF-8"> | |
| <meta name="viewport" content="width=device-width, initial-scale=1.0"> | |
| <title>mtDNA Tool – System Overview</title> | |
| <style> | |
| .custom-container { | |
| background-color: #ffffff ; | |
| color: #222222 ; | |
| font-family: Arial, sans-serif ; | |
| line-height: 1.6 ; | |
| padding: 2rem ; | |
| max-width: 900px ; | |
| margin: auto ; | |
| } | |
| .custom-container h1, | |
| .custom-container h2, | |
| .custom-container h3, | |
| .custom-container strong, | |
| .custom-container b, | |
| .custom-container p, | |
| .custom-container li, | |
| .custom-container ol, | |
| .custom-container ul, | |
| .custom-container span { | |
| color: #222222 ; | |
| font-weight: normal ; | |
| } | |
| .custom-container h1, | |
| .custom-container h2 { | |
| font-weight: bold ; | |
| } | |
| .custom-container img { | |
| max-width: 100%; | |
| border: 1px solid #ccc; | |
| padding: 5px; | |
| background: #fff; | |
| } | |
| .custom-container code { | |
| background: none ; | |
| color: #222 ; | |
| font-family: inherit ; | |
| font-size: inherit ; | |
| padding: 0 ; | |
| border-radius: 0 ; | |
| } | |
| .custom-container .highlight { | |
| background: #ffffcc; | |
| padding: 4px 8px; | |
| border-left: 4px solid #ffcc00; | |
| margin: 1rem 0; | |
| color: #333 ; | |
| } | |
| </style> | |
| </head> | |
| <body> | |
| <div class="custom-container"> | |
| <h1>mtDNA Location Classifier – Brief System Pipeline and Usage Guide</h1> | |
| <p>The <strong>mtDNA Tool</strong> is a lightweight pipeline designed to help researchers extract metadata such as geographic origin, sample type (ancient/modern), and optional niche labels (e.g., ethnicity, specific location) from mtDNA GenBank accession numbers. It supports batch input and produces structured Excel summaries.</p> | |
| <h2>System Overview Diagram</h2> | |
| <p>The figure below shows the core execution flow—from input accession to final output.</p> | |
| <img src="https://huggingface.co/spaces/VyLala/mtDNALocation/resolve/main/flowchart.png" alt="mtDNA Pipeline Flowchart"> | |
| <h2>Key Steps</h2> | |
| <ol> | |
| <li><strong>Input</strong>: One or more GenBank accession numbers are submitted (e.g., via UI, CSV, or text).</li> | |
| <li><strong>Metadata Collection</strong>: Using <code>fetch_ncbi_metadata</code>, the pipeline retrieves metadata like country, isolate, collection date, and reference title. If available, supplementary material and full-text articles are parsed using DOI, PubMed, or Google Custom Search.</li> | |
| <li><strong>Text Extraction & Preprocessing</strong>: | |
| <ul> | |
| <li>All available documents are parsed and cleaned (tables, paragraphs, overlapping sections).</li> | |
| <li>Text is merged into two formats: a smaller <code>chunk</code> and a full <code>all_output</code>.</li> | |
| </ul> | |
| </li> | |
| <li><strong>LLM-based Inference (Gemini + RAG)</strong>: | |
| <ul> | |
| <li>Chunks are embedded with FAISS and stored for reuse.</li> | |
| <li>The Gemini model answers specific queries like predicted country, sample type, and any niche label requested by the user.</li> | |
| </ul> | |
| </li> | |
| <li><strong>Result Structuring</strong>: | |
| <ul> | |
| <li>Each output includes predicted fields + explanation text (methods used, quotes, sources).</li> | |
| <li>Summarized and saved using <code>save_to_excel</code>.</li> | |
| </ul> | |
| </li> | |
| </ol> | |
| <h2>Output Format</h2> | |
| <p>The final output is an Excel file with the following fields:</p> | |
| <ul> | |
| <li><code>Sample ID</code></li> | |
| <li><code>Predicted Country</code> and <code>Country Explanation</code></li> | |
| <li><code>Predicted Sample Type</code> and <code>Sample Type Explanation</code></li> | |
| <li><code>Sources</code> (links to articles)</li> | |
| <li><code>Time Cost</code></li> | |
| </ul> | |
| <h2>System Highlights</h2> | |
| <ul> | |
| <li>RAG + Gemini integration for improved explanation and transparency</li> | |
| <li>Excel export for structured research use</li> | |
| <li>Optional ethnic/location/language inference using isolate names</li> | |
| <li>Quality check (e.g., fallback on short explanations, low token count)</li> | |
| <li>Report Button – After results are displayed, users can submit errors or mismatches using the report text box below the output table</li> | |
| </ul> | |
| <h2>Citation</h2> | |
| <div class="highlight"> | |
| Phung, V. (2025). mtDNA Location Classifier. HuggingFace Spaces. https://huggingface.co/spaces/VyLala/mtDNALocation | |
| </div> | |
| <h2>Contact</h2> | |
| <p>If you are a researcher working with historical mtDNA data or edge-case accessions and need scalable inference or logging, reach out through the HuggingFace space or email provided in the repo README.</p> | |
| </div> | |
| </body> | |
| </html> | |