UF Health Cancer Center Bioinformatics Hub
Experimental Design
RNA-seq

RNA-seq

Experimental Design Considerations

Replication Strategy

Plan for at least 3 biological replicates per condition, with 5-6 replicates preferred for detecting smaller effect sizes. Technical replicates are generally unnecessary with modern RNA-seq protocols.

Sample Collection and Storage

  • RNA integrity: Target RIN scores >7 for optimal results, though degraded samples (RIN 5-7) can still provide useful data with appropriate analysis adjustments
  • Storage: Flash freeze samples in liquid nitrogen and store at -80°C. Avoid freeze-thaw cycles
  • Batch effects: Randomize sample processing across batches when possible

Library Preparation Considerations

  • Poly(A) selection vs. rRNA depletion: Choose based on research goals - poly(A) for mRNA focus, rRNA depletion for comprehensive transcriptome including non-coding RNAs
  • Strand-specific libraries: Recommended for accurate transcript quantification and novel transcript discovery
  • PCR amplification: Minimize PCR cycles during library preparation to reduce duplication bias. Monitor final library complexity and duplication rates
  • Read length: 75-100bp paired-end reads are standard for most applications

Sequencing Depth

Target 20-30 million paired-end reads per sample for human/mouse samples. Increase depth for:

  • Novel transcript discovery (50+ million reads)
  • Low-input samples
  • Species with large genomes or high transcript diversity

Spike-in Controls

Consider ERCC spike-ins for:

  • Cross-sample normalization validation
  • Technical performance assessment
  • Absolute quantification needs

Quality Control Metrics

Monitor key metrics including:

  • Mapping rates (>80% for well-annotated genomes)
  • Duplication rates (<30% for high-quality libraries)
  • rRNA contamination (<10%)

ENCODE Guidelines

Follow ENCODE RNA-seq standards for experimental design and data processing. Key ENCODE benchmarks include >10 million uniquely mapped reads, <15% duplication rate, and strand-specificity >0.8 for stranded libraries.

Data Processing Workflow

[Link to analysis pipeline documentation]