Principles Behind RNA Sequencing

Background: Genetics & Gene Expression

• Genetics studies variation & heredity of phenotypic traits.
• Central dogma: $DNA \rightarrow mRNA \rightarrow Proteins$ (genes → transcripts → traits).
• Allele: specific nucleotide sequence at a locus; the combination of alleles = genotype.
• Phenotype is the observable trait (e.g., seed color/shape) resulting from genotype + gene regulation.
• Regulation layers influencing phenotype beyond Mendelian inheritance:
  • Alternative splicing
  • Non-coding RNAs
  • Gene networks & epigenetics

Genomics & Transcriptomics

• Genome = complete catalog of DNA variation.
• Transcriptomics measures expression of all genes simultaneously.
  • Two major technologies:
  • Microarrays (oligo-based, require prior sequence knowledge).
  • RNA-Seq (sequence-based, works even for unknown transcripts).
• Main drawback of microarrays: cannot detect transcripts lacking a pre-designed probe.

Historical Timeline of Expression Profiling

• $1960s$ – In-situ hybridization (probe staining of mRNA on slides).
• $1970s$ – Northern blots (size-separated RNA + probe on filters).
• $1985$ – RT-PCR amplification supersedes Northern blots; quantitative capability emerges.
• Late $1990s$ – Commercial microarrays (glass slides with bound oligos) debut.
• Early $2000s$ – Transgenic reporters to manipulate promoters.
• Mid-$2000s$ – Cost drop drives adoption of RNA-Seq; publications surge.

Experimental Objectives with RNA-Seq

• Quantitative questions:
  • Which genes/isoforms are expressed?
  • Relative abundance & differential expression across conditions.
  • Pathway enrichment correlations.
• Qualitative questions:
  • Novel transcripts & alternative start sites.
  • Strand origin of RNAs.
  • Sequence variants within expressed RNAs.
• Objective dictates library type, read length, & depth.

Library Preparation (recap)

1. Start with total RNA.
2. Fragment via heat to size-appropriate lengths.
3. Reverse-transcribe to cDNA using oligo(dT) or random primers.
4. Size-select with magnetic beads (uniform insert size).
5. Add 3′ $A$-tails.
6. Ligate adapters containing:
   • P5 & P7 flow-cell binding sites.
   • Index sequences (for multiplexing).
7. PCR-amplify to finalize library ready for sequencing.

Flow Cell & Cluster Generation (Illumina)

• Flow cell: glass slide with micro-channels coated by P5/P7 oligos.
• Bridge amplification steps:
  1. Single-stranded library molecule hybridizes to complementary oligo.
  2. Polymerase extends complementary strand.
  3. Original strand washed; new strand folds over to second oligo.
  4. Repeats → dense cluster (≈$1{-}2\ \mu m$) visible to camera.

Sequencing by Synthesis (Reversible Terminator Chemistry)

• Four fluorescently tagged, 3′-blocked dNTPs compete for incorporation.
• Cycle per base:
  1. Incorporate one nucleotide (polymerase).
  2. Image fluorescence (color ↔ base).
  3. Cleave dye & unblock 3′ end.
  4. Repeat.
• Colors: A=blue, C=orange, G=green, T=red (example mapping).
• Parallel imaging of $\sim10^8$ clusters ⇒ massive throughput.

Base Calling & Quality Scores

• For every cycle & cluster, intensity profile → base call.
• Q-score (Phred scale) reflects probability of incorrect call.
  • High Q = clear single-color peak.
  • Ambiguous intensities ⇒ call “N” (unknown) & low Q.

Single-End vs Paired-End Reads

• Single-end: sequence only one side of insert.
• Paired-end: sequence both ends; increases alignment confidence & detects indels.
• Instrument setup defines read length (e.g., $2\times75$ bp).

Indexing & Multiplexing

• Dual indices (i7/i5) embedded in adapters label sample origin.
• Enables pooling (e.g., lung, intestine, heart) in one lane; demultiplexing restores per-sample reads.

Primary Data Output

• Files: FASTQ (sequence + Q-scores).
• On-instrument pipeline: images → FASTQ = primary analysis.
• Secondary analysis: align to reference genome; tertiary: biological interpretation (pathways, DEGs).

Depth vs Coverage Concepts

• Read depth: number of reads mapping to a base.
• Breadth/Uniformity: proportion of reference spanned & evenness.
• Adequate depth crucial for:
  • Detecting low-abundance transcripts.
  • Confident variant calling.

Example thought question (seeded in lecture):

• Sample with $10\,\text{million}$ total reads.
  • Gene A: $8$ reads, length $=1\,kb$.
  • Gene B: $16$ reads, length $=2\,kb$.
• Raw counts alone cannot confirm higher expression; must normalize by transcript length (RPKM/TPM).

Sequencing Output Requirements

• Typical gene-expression mRNA-seq:
  • Stranded mRNA kit.
  • $\ge 75$ bp reads.
  • >10\,\text{million} reads/sample.
• Total RNA (rRNA-depleted):
  • $75{-}100$ bp reads.
  • >200\,\text{million} reads/sample.
• Choice of instrument (e.g., NovaSeq vs NextSeq) balances cost & throughput.

Sources of Bias & Variability

• Fragmentation bias: long transcripts generate more fragments → higher raw counts.
  • Remedy: normalize by length.
• Sampling bias: only subset of molecules enter library.
  • Low-expression genes quantified less accurately.
• Library-prep bias: adapter ligation efficiency, PCR duplicates.
• Sequencing bias: highly expressed transcripts saturate flow cell.
• Mitigations:
  • Increase depth, deplete abundant RNAs, correct computationally.
  • Employ biological replicates (outweigh technical replicates).

Technical & Biological Replicates

• Technical variation in RNA-seq generally low.
• Biological variation major driver → include multiple biological replicates to improve statistical power.

Sensitivity, Specificity, Dynamic Range

• Trade-offs summarized:
  • Sensitivity limited by read depth & starting RNA quality.
  • Specificity affected by ambiguous alignment; improved via longer/paired reads.
  • Dynamic range bounded by depth; highly expressed genes can mask rare ones.

Key Caveats & Best Practices

• Normalize raw counts by transcript length (RPKM/FPKM/TPM).
• Include adequate biological replicates.
• Scale sequencing depth/platform to experimental goals.
• Consider ribosomal depletion or poly-A enrichment to avoid saturation.
• Verify quality (Q-scores) & trim/adapt sequences before downstream analysis.

Summary Workflow Recap

1. Sample collection & RNA isolation.
2. Library prep (fragment, reverse-transcribe, adapter ligation, PCR).
3. Cluster generation on flow cell.
4. Sequencing by synthesis with reversible terminators.
5. Image acquisition → base calling → FASTQ (primary analysis).
6. Alignment & quantification (secondary).
7. Biological interpretation (tertiary): differential expression, pathways.

Illumina Video Highlights (supporting visualization)

• Demonstrates four stages: Sample Prep → Cluster Generation → Sequencing → Data Analysis.
• Explains bridge amplification, fluorescent imaging, index reads, paired-end turnaround, and final alignment to reference.

Final Take-Home Messages

• RNA-Seq provides comprehensive, high-resolution measurement of gene expression, surpassing microarrays.
• Success depends on thoughtful experimental design: library type, read length, depth, replicates, and bias mitigation.
• Understanding sequencing chemistry and data structure (FASTQ, Q-scores) is crucial for accurate downstream analysis.