Hugging Face's logo

Spaces:
VyLala
/
BioMetadataAudit
Running

App Files Community
1
BioMetadataAudit
File size: 4,133 Bytes 
e80d020
 
 
 
 
 
0ac4fe1
 
 
59fce1c
 
 
 
 
 
 
0ac4fe1
59fce1c
0ac4fe1
59fce1c
0ac4fe1
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
e80d020
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
 
<!DOCTYPE html>
<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>
  body {
    background-color: #ffffff !important;
    color: #222222 !important;
    font-family: Arial, sans-serif !important;
    line-height: 1.6 !important;
    padding: 2rem !important;
    max-width: 900px !important;
    margin: auto !important;
  }

  h1, h2 {
    color: #1a1a1a !important; /* darker headings */
  }
  img {
    max-width: 100%;
    border: 1px solid #ccc;
    padding: 5px;
    background: #fff;
  }
  code {
    background: #f5f5f5;
    color: #c7254e;
    padding: 2px 4px;
    border-radius: 4px;
    font-family: Consolas, monospace;
  }
  .highlight {
    background: #ffffcc;
    padding: 4px 8px;
    border-left: 4px solid #ffcc00;
    margin: 1rem 0;
    color: #333;
  }
  </style>
  
</head>
<body>

<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="./A_flowchart_in_the_image_illustrates_a_data_proces.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>

</body>
</html>