11.5 Genomics, Transcriptomics, and Proteomics

Core Explanation

The central dogma — DNA → RNA → protein — describes a unidirectional flow of information. But measuring each level reveals distinct, often discordant, views of the cell. Genomics tells us what the cell can do (its blueprint); transcriptomics tells us what it is trying to do (its active instructions); proteomics tells us what it is actually doing (its functional machinery). This topic covers the experimental approaches to each, their relative strengths, and — critically — why they do not always agree.

Genomics: The Blueprint

Technologies

Next-Generation Sequencing (NGS) has largely replaced Sanger sequencing for genome-scale projects. The dominant platform (Illumina) uses sequencing-by-synthesis with reversible dye terminators: fluorescently labeled, 3′-blocked dNTPs are incorporated one base at a time, imaged, and cleaved to allow the next cycle. This generates millions of short reads (50–300 bp) in parallel.

Third-generation sequencing (PacBio SMRT, Oxford Nanopore) produces long reads (10–100+ kb) that span repetitive regions and structural variants inaccessible to short-read platforms.

Applications

Key Limitations

• Static: The genome is (mostly) the same in every cell of an organism throughout its life. Genomics reveals potential, not current activity.
• Non-coding interpretation: >98% of the human genome is non-coding; assigning function to regulatory elements remains an enormous challenge.

Transcriptomics: The Active Instructions

RNA-Seq Workflow

Step 1: RNA Extraction Total RNA is extracted and assessed for integrity (RNA Integrity Number, RIN). Even modest degradation introduces 3′ bias and reduces detection of low-abundance transcripts.

Step 2: Enrichment/Depletion

• Poly-A selection: Oligo(dT) beads capture mRNA via poly-A tails. Enriches for mRNA but misses non-polyadenylated transcripts (histone mRNAs, some lncRNAs).
• rRNA depletion: Ribosomal RNA (>80% of total RNA) is selectively removed, retaining all non-ribosomal species. Preferred for degraded samples (e.g., FFPE tissue) and lncRNA discovery.

Step 3: Library Preparation RNA is fragmented, reverse-transcribed to cDNA, and adapters are ligated. For strand-specific libraries, the second strand is marked (e.g., with dUTP), enabling unambiguous determination of which DNA strand was the original template — critical for distinguishing overlapping sense and antisense transcripts.

Step 4: Sequencing Libraries are sequenced to a depth appropriate for the application: 20–50 million reads/sample for standard differential expression; 50–100 million for transcript assembly and isoform analysis.

Step 5: Alignment and Quantification Reads are aligned to a reference genome (STAR, HISAT2) or pseudoaligned to a transcriptome (Salmon, kallisto). Expression is quantified as counts per gene or transcript, then normalized to account for library size and gene length:

• TPM (Transcripts Per Million): Normalizes for gene length and library size; values across samples are comparable.
• FPKM/RPKM: Similar to TPM; TPM is generally preferred for cross-sample comparison.

Step 6: Differential Expression Analysis Tools (DESeq2, edgeR, limma-voom) model count data using a negative binomial distribution and identify genes with statistically significant changes between conditions. Results are typically visualized as:

• Volcano plot: log₂(fold change) on x-axis vs. −log₁₀(adjusted p-value) on y-axis. Significantly up- and down-regulated genes appear at the extremes.
• Heatmap: Clustering of samples and genes reveals patterns of co-regulation.

Key Limitation: RNA Abundance ≠ Protein Abundance

This is one of the most important — and frequently ignored — facts in molecular biology. Genome-wide studies consistently find that mRNA levels explain only ~30–40% of the variance in protein abundance (Schwanhäusser et al., 2011; Vogel & Marcotte, 2012). The intervening factors include:

1. Translation rate: Ribosome occupancy varies across transcripts; some mRNAs are efficiently translated, others are sequestered in P-bodies or stress granules.
2. Protein half-life: Proteins vary dramatically in stability — from minutes (cyclins, c-Fos) to months (crystallins in the lens). A stable protein at low mRNA can be more abundant than an unstable protein at high mRNA.
3. Post-translational regulation: Ubiquitination, autophagy, and proteasomal degradation dynamically modulate protein levels independent of synthesis.
4. Cellular heterogeneity: Single-cell RNA-Seq reveals that mRNA levels vary widely among individual cells; bulk transcriptomics averages this out, masking subpopulations.

Practical rule: Transcriptomics identifies candidate genes. Proteomics (or at minimum, Western blot validation) is required to confirm that changes in mRNA translate to changes in protein.

Proteomics: The Functional Machinery

Proteomics measures the complete complement of proteins — their identities, abundances, modifications, interactions, and localizations. It is fundamentally more challenging than genomics or transcriptomics because:

1. No amplification: Unlike DNA/RNA, proteins cannot be amplified. Detection must work with whatever is present. This makes low-abundance proteins (transcription factors, signaling kinases) notoriously difficult to detect.
2. Dynamic range: Protein abundances span >7 orders of magnitude in human cells (albumin in plasma: ~40 mg/mL; cytokines: ~pg/mL). No single method covers this range.
3. Chemical diversity: The 20 amino acids exhibit vastly different chemical properties (charge, hydrophobicity, size), and post-translational modifications (PTMs) add hundreds of additional forms. No single separation or detection method works for all proteins.
4. Splice isoforms and proteoforms: One gene can produce multiple protein isoforms via alternative splicing, PTMs, and proteolytic processing. Each proteoform may have distinct function.

Mass-Spectrometry-Based Proteomics Workflow

Step 1: Protein Extraction and Digestion Proteins are extracted under denaturing conditions, reduced (DTT/TCEP) and alkylated (iodoacetamide) to block cysteine residues, and digested with trypsin — a protease that cleaves C-terminally to lysine and arginine, producing peptides of predictable masses ideal for MS analysis.

Step 2: Peptide Separation Peptides are separated by reversed-phase liquid chromatography (LC), typically with a gradient of increasing acetonitrile, which elutes peptides by hydrophobicity. Multidimensional separation (e.g., strong cation exchange + reversed phase) increases coverage.

Step 3: Mass Spectrometry (MS) Eluting peptides are ionized (electrospray ionization, ESI) and introduced into the mass spectrometer. Two stages:

• MS1 (survey scan): Measures the mass-to-charge (m/z) ratios of intact peptide ions. The most abundant ions are selected for fragmentation.
• MS2 (tandem MS / MS/MS): Selected precursor ions are fragmented (collision-induced dissociation, CID) and the fragment masses are recorded. The fragmentation pattern serves as a peptide "fingerprint."

Step 4: Peptide Identification MS/MS spectra are searched against a protein sequence database using algorithms (Mascot, SEQUEST, MaxQuant/Andromeda) that perform in silico digestion and match experimental spectra to theoretical fragmentation patterns. False discovery rate (FDR) is controlled by searching against a decoy (reversed) database; a 1% FDR at the peptide level is standard.

Step 5: Quantification

Post-Translational Modification (PTM) Analysis

PTMs (phosphorylation, ubiquitination, acetylation, glycosylation, etc.) are identified by characteristic mass shifts in MS/MS spectra. Enrichment strategies (e.g., immobilized metal affinity chromatography for phosphopeptides) are essential because modified peptides are often substoichiometric and invisible in unfractionated samples. Phosphoproteomics routinely identifies >10,000 phosphorylation sites, but assigning function to individual sites remains a major bottleneck.

Comparing the Three Omics

Readout

• Genomics: VCF files (variants), BAM files (alignments), genome browsers (IGV).
• Transcriptomics: Count matrices, volcano plots, MA plots, PCA plots (quality control), heatmaps, gene set enrichment analyses.
• Proteomics: Peptide-spectrum matches, protein groups tables, volcano plots, PTM occupancy ratios.

Strengths

• Integration: Combined multi-omics provides a systems-level view unattainable from any single layer.
• Discovery-based: Unlike hypothesis-driven methods (qPCR, Western), omics approaches are unbiased and can reveal unexpected players.
• Quantitative: All three can be quantitative, enabling statistical comparison across conditions.

Common Interpretation Errors

• "My RNA-Seq shows Gene X is upregulated 5-fold, so the protein must be upregulated too." mRNA-protein correlation is modest. A 5-fold mRNA change may produce no protein change, or vice versa. Transcriptomic results are hypotheses; proteomic validation is evidence.
• "The most differentially expressed gene is the most important." Statistical significance (p-value) is driven by effect size AND variance. A gene with a 1.5-fold change and low variance can be more "significant" than a 10-fold change with high variance. Biological importance and statistical significance are not the same.
• "My proteomics experiment identified 3,000 proteins, so I'm missing the other ~7,000." Coverage depends on sample complexity, fractionation, and instrument sensitivity. Missing proteins are not necessarily absent — they may be below the detection limit, poorly ionized, or obscured by highly abundant proteins. A negative proteomics result is rarely proof of absence.
• "The volcano plot threshold (log₂FC > 1, p < 0.05) defines all the 'real' hits." Arbitrary cutoffs discard biologically meaningful changes with small effect sizes (e.g., transcription factors where 2-fold is biologically significant) and may include statistically significant but biologically trivial changes.

Quick Questions

Q1: You perform RNA-Seq on cells treated with a drug and find that Gene W is upregulated 8-fold (adjusted p = 10⁻¹²). You run a Western blot and see no change in Gene W protein. List three biologically plausible explanations.

Q2: Why is proteomics coverage typically lower than transcriptomics coverage, and what strategies can improve it?

Q3: What does a volcano plot show, and why might a gene with a 10-fold change not appear as a "hit" while a gene with a 1.5-fold change does?