Map the major -omics types onto the spine

By now you can follow one molecule all the way down the spine. DNA in S3, transcribed to RNA in S6, translated to protein in S5, the whole flow switched on and off by the regulation of S8. And in BIO-4.1 you met the machine that turns any of these molecules into data: a sequencer that emits short reads with known biases. This lesson zooms out. It names the major -omics, the family of comprehensive measurements biologists take, and shows that each one is just a different probe pointed at a layer of the spine you already own.

Start with the word. The suffix -ome means "the complete set of." Your genome is the complete set of your DNA. Your transcriptome is the complete set of RNA molecules in a sample right now. Add -ics and you get the study and measurement of that set. So -omics is not new biology. It is a promise to measure something comprehensively instead of one gene at a time, and the something is always a layer you have already met.

Here is the unifying trick, and the first surprise. Three of the four big -omics run on the same hardware from BIO-4.1. If a sequencer can only read A, C, G, T, then to measure RNA you first copy it back into DNA, and to read DNA methylation you chemically convert the genome so the marks show up as different letters. Almost everything becomes a sequencing problem. Proteomics is the exception, and its exception is exactly why it is the hardest.

Genomics: the blueprint that barely changes

Genomics is the -omics of DNA itself: the sequence and its variants (recall the variant menu from S7.1, substitutions, indels, and larger structural changes). You sequence a genome, map the reads back to a reference with the alignment machinery from BIO-3, and ask where this sample differs from that reference.

What it answers: which version of the code you carry. Which alleles, which disease variants, which mutations a tumor has accumulated. Because every cell in your body carries essentially the same DNA (the load-bearing fact from S8.1), one blood sample reports the genome of your whole body, give or take somatic mutations.

What it cannot answer: whether any of that code is being used. A gene can sit in your genome, perfectly intact, and never once be expressed in the cell you sampled. Genomics reads the blueprint. It says nothing about what the factory is building today.

Transcriptomics: what the cell is expressing right now (RNA-seq)

Transcriptomics measures the transcriptome, the population of RNA molecules present at the moment you freeze the sample. The dominant method is RNA-seq. You extract RNA, reverse-transcribe it into complementary DNA (cDNA) so the sequencer can read it, sequence the lot, and then count: for each gene, how many reads came from its transcripts. More reads means more of that RNA was present, which you read as higher expression.

This is the biology of S8 made quantitative. S8.1 told you cell types differ not by their DNA but by which genes they express and how much. RNA-seq is how you actually see that difference. A liver cell and a neuron share one genome and carry wildly different transcriptomes, and the transcriptome is what the cell is doing, not just what it could do.

What it answers: which genes are on, and roughly how loud, in this sample, at this moment. Differential expression (this gene is 8 times higher in tumor than in healthy tissue) is the workhorse result.

What it cannot answer, and this is the trap the whole lesson is circling: how much protein there is.

Make the gap concrete. Read this panel and predict the printed ratios before you trace it. The numbers are illustrative, not a live run.

# mRNA counts come from RNA-seq. protein comes from mass spec.
# Made-up but realistic numbers.
measured = {
    "geneA": {"mrna": 100, "protein": 5000},
    "geneB": {"mrna": 100, "protein": 400},
    "geneC": {"mrna": 20,  "protein": 6000},
}

# If mRNA fully determined protein, protein / mrna would be one
# constant for every gene. Watch it refuse to be.
for gene, m in measured.items():
    ratio = m["protein"] / m["mrna"]
    print(gene, "protein per mRNA =", ratio)

geneA and geneB have identical mRNA and very different protein. geneC has the least mRNA and the most protein. There is no constant you can multiply an RNA count by to recover a protein count. That missing constant is everything S8.4 controls: how each transcript is translated, degraded, and destroyed.

Proteomics: the actual workers (mass spectrometry)

Proteomics measures the proteome, the actual set of proteins present, and it is the direct measurement that transcriptomics only approximates. But proteins are not a four-letter chain you can copy and feed to a sequencer, so the sequencing hardware is useless here. The main tool is mass spectrometry: you chop proteins into peptide fragments, ionize them, and weigh those fragments with extreme precision, then match the measured masses back against the protein sequences (predicted, recall, from the genome through the codon table of S6) that could have produced them.

The catch is that there is no PCR for proteins. DNA and RNA can be amplified, copied billions of times, so even a trace becomes detectable. A protein present in a few copies cannot be photocopied, so proteomics is systematically blind to low-abundance proteins and covers its layer far less completely than RNA-seq covers the transcriptome. The layer closest to what a cell actually does is also the hardest and least complete to read. That gap, where more protein data would settle questions mRNA only hints at, is a live frontier, not a solved problem.

Epigenomics: the settings on top of the sequence

Epigenomics measures the marks and physical states layered on the DNA without changing its sequence, straight out of S8.3: methylation (a chemical tag added to cytosine bases) and chromatin accessibility (whether a stretch of DNA is open for the machinery to reach or packed away out of reach). Methylation is read by bisulfite sequencing, a chemistry that converts unmethylated cytosine so marked and unmarked bases read as different letters on the very same sequencer from BIO-4.1. Accessibility is read by assays like ATAC-seq, which cut and sequence only the open regions.

What it answers: not the code, but the settings on the code. Which regions are switched into a permissive (on) or silenced (off) configuration, why two cells with identical DNA behave differently, and how those settings drift with age.

Single-cell: from the smoothie to the individuals

One more axis cuts across all of the above: whether you measure a bulk sample or single cells. Classic RNA-seq is bulk. You take a chunk of tissue, millions of cells, grind them together, and read the average. The trouble with an average is the trouble with every average. Blend a smoothie of a thousand cells and you cannot tell whether every cell held medium expression of a gene, or half the cells screamed it while half were silent. The two blends taste identical.

Single-cell methods (single-cell RNA-seq is the common one) tag each cell's molecules with a unique barcode before pooling, so after sequencing you can sort the reads back to the individual cell they came from. Now instead of one average you get a distribution: this tissue is actually five cell types in these proportions, and the "medium" bulk signal was really two populations moving in opposite directions.

Put it together and the map is simple. The genome is what you could do, and it barely changes across your cells. The transcriptome is what a cell is expressing right now, measured by counting RNA reads, and it is a proxy for, not a measurement of, the protein layer. The proteome is what is actually getting done, measured directly but incompletely. The epigenome is the settings that decide which of the genome is reachable in the first place. And bulk versus single-cell decides whether you see the average or the individuals. Every one of these is the same spine, read at a different layer, with a different instrument, carrying a different blind spot you now know to name.