Skip to content

PIPELINES.md

Latest commit

 

History

History
203 lines (152 loc) · 4.34 KB

PIPELINES.md

File metadata and controls

203 lines (152 loc) · 4.34 KB

RAPTOR Pipelines Reference

Deep dive into all 8 RNA-seq analysis pipelines.

Pipeline Overview

ID Name Type Speed Accuracy Memory
1 STAR-RSEM-DESeq2 Alignment ⚫⚫⚪⚪⚪ ⚫⚫⚫⚫⚫ High
2 HISAT2-StringTie-Ballgown Alignment ⚫⚫⚫⚪⚪ ⚫⚫⚫⚫⚪ Medium
3 Salmon-edgeR Pseudo-align ⚫⚫⚫⚫⚫ ⚫⚫⚫⚫⚪ Low
4 Kallisto-Sleuth Pseudo-align ⚫⚫⚫⚫⚫ ⚫⚫⚫⚪⚪ Low
5 STAR-HTSeq-limma Alignment ⚫⚫⚪⚪⚪ ⚫⚫⚫⚫⚪ High
6 STAR-featureCounts-NOISeq Alignment ⚫⚫⚪⚪⚪ ⚫⚫⚫⚪⚪ High
7 Bowtie2-RSEM-EBSeq Alignment ⚫⚫⚪⚪⚪ ⚫⚫⚫⚪⚪ Medium
8 HISAT2-Cufflinks-Cuffdiff Alignment ⚫⚪⚪⚪⚪ ⚫⚫⚪⚪⚪ Medium

Pipeline 1: STAR-RSEM-DESeq2

Gold Standard - Highest Accuracy

Components

  • Alignment: STAR (splice-aware)
  • Quantification: RSEM (EM algorithm)
  • Statistics: DESeq2 (negative binomial)

Best For

  • Publication-quality results
  • Small sample sizes (n<6)
  • Complex experimental designs
  • When accuracy is paramount

Performance

  • Runtime: ~2-6 hours (typical dataset)
  • Memory: 32-48 GB
  • Accuracy: 95%

Running

raptor run --pipeline 1 \
  --data fastq/ \
  --reference /path/to/star_index \
  --annotation genes.gtf \
  --output pipeline1_results/

Pipeline 2: HISAT2-StringTie-Ballgown

Transcript Assembly & Novel Discovery

Components

  • Alignment: HISAT2
  • Assembly: StringTie
  • Statistics: Ballgown

Best For

  • Novel transcript discovery
  • Isoform-level analysis
  • Non-model organisms
  • When reference incomplete

Performance

  • Runtime: ~1-4 hours
  • Memory: 16-24 GB
  • Accuracy: 88%

Pipeline 3: Salmon-edgeR ⭐ RECOMMENDED

Best Balance - Fast & Accurate

Components

  • Quantification: Salmon (quasi-mapping)
  • Statistics: edgeR (quasi-likelihood)

Best For

  • Most RNA-seq experiments
  • Large datasets (>20 samples)
  • Quick turnaround needed
  • Good balance of all metrics

Performance

  • Runtime: ~0.5-2 hours
  • Memory: 8-16 GB
  • Accuracy: 90%

Running

raptor run --pipeline 3 \
  --data fastq/ \
  --transcriptome /path/to/salmon_index \
  --output pipeline3_results/

Pipeline 4: Kallisto-Sleuth

Ultra-Fast - Large Studies

Components

  • Quantification: Kallisto
  • Statistics: Sleuth (bootstrap-based)

Best For

  • Very large datasets (>50 samples)
  • Exploratory analysis
  • Minimal resources
  • Speed is critical

Performance

  • Runtime: ~0.3-1 hour
  • Memory: 4-8 GB
  • Accuracy: 88%

Pipeline 5: STAR-HTSeq-limma-voom

Flexible Modeling - Complex Designs

Components

  • Alignment: STAR
  • Counting: HTSeq
  • Statistics: limma-voom

Best For

  • Complex experimental designs
  • Multi-factor analysis
  • Batch correction needed
  • Repeated measures

Performance

  • Runtime: ~2-7 hours
  • Memory: 32-40 GB
  • Accuracy: 92%

Pipeline 6: STAR-featureCounts-NOISeq

Non-Parametric - Small Samples

Components

  • Alignment: STAR
  • Counting: featureCounts
  • Statistics: NOISeq

Best For

  • Very small samples (n=2-3)
  • No replicates
  • Non-normal distributions

Performance

  • Runtime: ~2-7 hours
  • Memory: 32-36 GB
  • Accuracy: 85%

Pipeline 7: Bowtie2-RSEM-EBSeq

Memory-Efficient Alternative

Components

  • Alignment: Bowtie2
  • Quantification: RSEM
  • Statistics: EBSeq

Best For

  • Moderate resource environments
  • Isoform-level analysis
  • Two-condition comparisons

Performance

  • Runtime: ~3-8 hours
  • Memory: 16-24 GB
  • Accuracy: 87%

Pipeline 8: HISAT2-Cufflinks-Cuffdiff

Legacy Pipeline

Components

  • Alignment: HISAT2
  • Assembly: Cufflinks
  • Statistics: Cuffdiff

Best For

  • Reproducing legacy analyses
  • When Cufflinks ecosystem required

Performance

  • Runtime: ~4-12 hours
  • Memory: 20-32 GB
  • Accuracy: 82%

Note: Newer methods preferred for new projects

Choosing a Pipeline

Use the decision guide:

  1. Need highest accuracy? → Pipeline 1
  2. Large dataset? → Pipeline 3 or 4
  3. Novel transcripts? → Pipeline 2
  4. Complex design? → Pipeline 5
  5. Small samples? → Pipeline 6
  6. Not sure? → Use raptor profile

See configuration details in config/pipelines.yaml