rRNA NOR Pipeline

A computational pipeline for analyzing 45S ribosomal RNA gene sequences from the Nucleolus Organizer Region (NOR) of plant genomes, with cross-species comparative analysis using machine learning.

Developed as part of a Study Oriented Project at BITS Pilani, Hyderabad Campus.

---

What this pipeline does

Plant genomes contain hundreds of tandemly repeated 45S rRNA gene copies clustered in the NOR region of specific chromosomes. These copies are not identical — they vary in sequence, contain insertions of different sizes, and can be grouped into haplotypes. This pipeline provides tools to:

• Load and process multiple MAFFT-aligned FASTA files from rDNA NOR regions
• Extract k-mer frequency features and gap pattern features from aligned sequences
• Compute sequence similarity to reference species (downloads automatically from NCBI)
• Cluster sequences using K-Means and compare clustering consistency across alignments
• Classify insertion types using a Random Forest classifier
• Visualize sequence diversity using UMAP dimensionality reduction
• Generate conservation heatmaps comparing Micro-Tom to reference species
• Analyze the spatial organisation of haplotypes across the NOR chromosomal region
• Characterize specific insertion sequences: BLAST search, TE analysis, regulatory motif scanning

---

Pipeline overview

Your FASTA files          Reference sequences (NCBI)
      │                           │
      └──────────┬────────────────┘
                 ↓
        Feature extraction
      k-mer vectors + gap vectors
                 ↓
      ┌──────────┴──────────┐
      ↓                     ↓
  K-Means               Random Forest
  clustering            gap classifier
      ↓                     ↓
  UMAP plot         Gap importance plot
      ↓
  Conservation heatmap
      ↓
  Positional NOR map
      ↓
  Summary report (CSV)

---

Repository structure

rRNA-NOR-pipeline/
├── README.md
├── requirements.txt
├── LICENSE
├── .gitignore
│
├── pipeline/
│   ├── rRNA_pipeline.py       # Main ML pipeline (Steps 1-12)
│   ├── analyze_insertion.py   # Insertion sequence analysis (BLAST, TE, regulatory)
│   └── haplotype_analysis.R   # R scripts for haplotyping (ape, pegas, geneHapR)
│
├── demo/
│   ├── generate_demo_data.py  # Generates synthetic FASTA files for testing
│   └── demo_sequences.fasta   # Pre-generated synthetic demo data
│
└── docs/
    ├── methods.md             # Full step-by-step methodology
    └── pipeline_overview.png  # Visual pipeline diagram

---

Installation

Requirements: Python 3.8 or higher, Windows/Mac/Linux

Step 1 — Clone the repository

git clone https://github.com/YOUR_USERNAME/rRNA-NOR-pipeline.git
cd rRNA-NOR-pipeline

Step 2 — Install dependencies

pip install -r requirements.txt

---

Quick start with demo data

To verify the pipeline works on your system before using your own data:

# Generate synthetic demo sequences
python demo/generate_demo_data.py

# Run the full pipeline on demo data
python pipeline/rRNA_pipeline.py --demo

This will create a pipeline_output/ folder with all figures and reports generated from synthetic sequences.

---

Usage with your own data

Step 1 — Prepare your files

Place your MAFFT-aligned FASTA files in a single folder. The pipeline expects:

• One or more alignment files (e.g. Alignment_1.fasta, Alignment_2.fasta)
• Optionally: haplotype summary files

Sequence IDs should ideally encode chromosomal position in the format:

NOR_region/START_END

for example: NOR2_region_(15.8_Mb)/8825_17270

This allows the pipeline to generate the positional NOR map (Step 9).

Step 2 — Edit the configuration

Open pipeline/rRNA_pipeline.py and edit the two lines at the top:

ALIGNMENT_DIR = r"C:\path\to\your\fasta\folder"   # your folder
YOUR_EMAIL    = "your_email@example.com"            # for NCBI API

Step 3 — Run

python pipeline/rRNA_pipeline.py

Step 4 — Run insertion analysis (optional)

If you have a FASTA file containing specific insertion sequences to characterize:

# Edit FASTA_141BP path in analyze_insertion.py first
python pipeline/analyze_insertion.py

Step 5 — Run haplotype analysis in R (optional)

Open pipeline/haplotype_analysis.R in RStudio. Three approaches are available:

# Method 1 — standard ape/pegas haplotyping
hap1 <- method1_haplotyping("path/to/alignment.fasta")

# Method 2 — geneHapR with gap filtering
hap2 <- method2_haplotyping("path/to/alignment.fasta")

# Find optimal number of clusters first
optimal_k <- find_optimal_clusters("path/to/alignment.fasta")

# K-Means sequence clustering
km <- cluster_sequences("path/to/alignment.fasta", n_clusters = optimal_k)

---

Output files

The pipeline generates a pipeline_output/ folder containing:

File	Description
step7_umap.png	UMAP of all sequences colored by cluster, reference species as stars
step8_conservation_heatmap.png	Conservation along gene body vs reference species
step9_positional_map.png	Cluster distribution across NOR chromosomal region
step10_gap_importance.png	Random Forest feature importance for insertion classification
step11_haplotype_similarity.csv	Per-haplotype similarity to each reference species
step12_summary_report.csv	Cross-alignment comparison table

For insertion analysis:

File	Description
step1_conservation_profile.png	Per-position conservation, GC content, dinucleotide heatmap
step2_blast_results.png	BLAST hits ranked by identity and significance
step2_blast_summary.csv	Full BLAST results table
step3_te_analysis.png	Transposable element signature analysis
step4_regulatory_analysis.png	Regulatory motif positions and prevalence
step5_summary.txt	Plain text summary with biological interpretation

---

Reference species used

The pipeline automatically downloads these sequences from NCBI:

Species	Accession	Role
Arabidopsis thaliana	X52322	Well-annotated reference
Solanum lycopersicum SL4.0	AY305797	Closest relative to Micro-Tom
Oryza sativa	AY373817	Distant outgroup

---

Methods summary

Feature extraction — Each sequence is converted to a normalised k-mer frequency vector (k=4, 256 features) capturing sequence composition, and a binary gap pattern vector (200 bp windows) capturing insertion presence/absence.

Clustering — K-Means clustering (k=9) is applied to the combined feature matrix. Adjusted Rand Index is computed between alignments to assess consistency.

Gap classification — A Random Forest classifier (300 trees) is trained on gap vectors to predict insertion type. Feature importance identifies which chromosomal positions best distinguish insertion variants.

Reference similarity — Cosine similarity between k-mer vectors measures sequence divergence from reference species without requiring a new alignment.

Dimensionality reduction — UMAP projects sequences into 2D space for visualization, embedding Micro-Tom and reference sequences together for direct comparison.

---

Citation

If you use this pipeline in your work, please cite:

Dhamele, A. (2025). rRNA NOR Pipeline:
GitHub: https://github.com/YOUR_USERNAME/rRNA-NOR-pipeline
BITS Pilani, Hyderabad Campus.

---

⭐ Found this useful?

If this pipeline helped your research, please consider starring the repo — it helps others discover it and lets me know the work is being used!

Author

Aarati Dhamele
B.E. Computer Science+ M.Sc Biological Sciences
BITS Pilani, Hyderabad Campus
ID: 2023B1A70641H

---

Acknowledgements

This pipeline was developed under the Study Oriented Project program at BITS Pilani Hyderabad. Reference sequences obtained from NCBI GenBank. Pipeline uses Biopython, scikit-learn, UMAP-learn, matplotlib, and seaborn.