flowchart TD
A[Raw signal / FAST5] --> B[Basecalling]
B --> C[FASTQ reads]
C --> D[QC: Q score / read length]
D --> E{QC pass?}
E -->|No| F[Re-extract / clean-up / re-sequence]
E -->|Yes| G[Kraken2 classification]
G --> H{Candidate pathogen?}
H -->|No| I[Check database / contamination / depth]
H -->|Yes| J[minimap2 alignment]
J --> K[Coverage / depth / mapping quality]
K --> L{Sufficient evidence?}
L -->|No| M[Increase depth / refine reference]
L -->|Yes| N[Biological interpretation]
N --> O[Report / decision making]
Nanopore analysis workflow
End-to-end pipeline for pathogen detection and interpretation
Note
Public This page provides a complete end-to-end Nanopore analysis workflow, from raw data to biological interpretation.
Overview
Nanopore sequencing 的分析流程可以拆成四個核心層級:
- Data quality(資料品質)
- Composition(樣本組成)
- Validation(結果驗證)
- Interpretation(生物意義)
👉 每一層回答不同問題,缺一不可
Full workflow
Step-by-step explanation
1️⃣ Basecalling
- 將 raw signal(FAST5)轉成 FASTQ
- 工具:Guppy / Dorado
👉 這一步決定後續所有分析品質
2️⃣ QC(Quality control)
對 FASTQ 做基本評估:
- Q score(品質)
- Read length / N50
- Total bases
- Read count
👉 對應頁面:
3️⃣ Classification(Kraken2)
目的:
找出樣本中「可能存在的生物」
特點:
- 快速
- 高通量
- 適合初步篩檢
👉 對應頁面:
4️⃣ Alignment(minimap2)
目的:
驗證特定目標是否存在
評估指標:
- coverage
- depth
- mapping quality
- read consistency
👉 對應頁面:
5️⃣ Interpretation
這一步最關鍵,也是最容易被忽略:
需要整合:
- read count
- coverage
- negative control
- biological plausibility
- sample context
👉 這一步沒有單一工具可以取代
Key decision points
QC 不佳
👉 問題可能在:
- DNA degradation
- extraction method
- library prep
- contamination
👉 不建議直接往下分析
Kraken2 有 hits
👉 不代表一定存在!
需要:
- minimap2 驗證
- 看 coverage(不是只看 read 數)
minimap2 有 alignment
👉 要確認:
- 是否為特異性 mapping
- 是否有均勻 coverage
- 是否可能為 contamination
Minimal working pipeline(實務版)
# 1. QC
NanoPlot --fastq reads.fastq -o qc_report
# 2. classification
kraken2 --db kraken_db reads.fastq --report report.txt > output.txt
# 3. alignment
minimap2 -ax map-ont ref.fa reads.fastq > aln.samCommon pitfalls
Warning
Nanopore workflow 最常見錯誤來自「過度簡化分析流程」。
- 只做 Kraken2 就下結論
- 忽略 QC
- 不看 coverage
- 不設 negative control
- 過度解讀低 read count
Best practice summary
Tip
一個可靠的 Nanopore 分析流程應該同時具備:
- QC(確保資料可信)
- Classification(找方向)
- Alignment(確認證據)
- Interpretation(整合判讀)
Suggested figures(極需)
🖼️ Figure 1(已經有了)
👉 Mermaid workflow(上面那張)
🖼️ Figure 2(未來可以加)
👉 真實案例:
- Kraken2 barplot
- minimap2 coverage plot
👉 提升網站變「教學等級」