Biology › Section I · Biotechnology and genomics › Topic I.3
Applying genomics and proteomics
Sequencing a genome is no longer the hard part; using that information well is. This topic surveys what whole genomes, transcriptomes and proteomes are teaching medicine, agriculture and forensics, and the ethical questions that follow.
From sequence data to decisions
Whole-genome sequencing turned biology into an information science: once reading DNA became fast and cheap, researchers began comparing thousands of genomes at once, and the payoff reaches far outside the laboratory. Doctors want to predict disease before symptoms appear, breeders want the genes behind drought tolerance, and investigators have traced an outbreak to its source by genomics. This topic follows that pipeline: how genome data become risk estimates, how the expressed genes and proteins of a cell are measured, and what safeguards the data need.
Predicting disease risk
Screening a healthy person's genome can reveal variants that raise the risk of future disease, so lifestyle changes or medication can start before onset. But the approach has a hard limit: single-gene defects, where one mutation decides the outcome, account for only about 5% of disease in developed countries. The heavy hitters, heart disease, type 2 diabetes and most cancers, are polygenic: many genes contribute small effects alongside environment. So a genome mostly reports probabilities, not verdicts. Genome-wide association studies (GWAS) meet this by scanning thousands of genomes and asking which variants appear more often in people with a disease than in matched healthy controls.
Worked example: reading a GWAS result
A GWAS compares 2,000 patients with type 2 diabetes and 2,000 healthy controls. A variant near a gene involved in insulin secretion appears in 32% of patients (640 people) but only 24% of controls (480 people). Reading it in steps: (1) The variant is genuinely more common among patients, a gap of 8 percentage points across thousands of people, and a statistical test confirms the gap is very unlikely to be chance. (2) Association is not causation: the variant may only sit near a causal gene, tagging it because nearby DNA is inherited together. (3) The effect is modest: most people carrying the variant never develop diabetes, and most patients do not carry it. (4) Its value is a pointer: a gene and pathway worth studying with experiments.
Transcriptomics: which genes are on
The genome is nearly identical in every cell, but gene expression is not. Transcriptomics measures which genes a sample is actually reading, by sequencing its RNA (RNA-seq) or by washing the sample over a DNA microarray, a slide carrying hundreds of thousands of known probes. Comparing a tumour with healthy tissue from the same organ shows which genes are switched up or down in the cancer. Single-cell RNA-seq refines the picture further by measuring one cell at a time, revealing that cells in one tissue can differ dramatically, a discovery hidden in bulk averages.
Proteomics: what the cell actually built
Genes and transcripts are only proxies: proteins do most of the work. The proteome is the complete set of proteins a cell type produces, and it is not a mirror of its mRNA counts. Some transcripts are never translated; others are alternately spliced, so one gene can yield several proteins; and proteins are chemically modified after translation, by phosphate groups, sugar chains or cutting. Mass spectrometry is the workhorse: it breaks proteins into fragments, weighs each fragment precisely, and matches the pattern against known proteins. The genome is constant while the proteome shifts hour by hour, so the proteome reports what a cell is doing, not merely what it could do.
Worked example: a protein signature
Researchers want a blood test that flags early pancreatic cancer, a disease usually caught late. They run mass spectrometry on 100 blood samples from early-stage patients and 100 from healthy donors, and find a set of four proteins that is high in the patients. Evaluating in steps: (1) The signature correctly flags 88 of the 100 patient samples, a sensitivity of 88%. (2) It also alarms in 6 of the 100 healthy samples, a 6% false-positive rate, so about 1 in 17 healthy people screened would be wrongly told they might have cancer. (3) For a cancer affecting 1 in 10,000 people, most positive results would be false alarms, so rarity matters. (4) A signature must be validated on large independent groups before screening programmes adopt it.
Systems biology and structure prediction
Genes and proteins act in networks, and systems biology studies those networks rather than single parts. Interaction maps show hub proteins that touch dozens of partners; disturbing a hub tends to harm the cell more than disturbing a peripheral protein, one reason hubs are often drug targets. The same network mindset, plus machine learning, cracked a 50-year-old problem: predicting a protein's three-dimensional shape from its amino-acid sequence. The AlphaFold system, announced in 2020, reached accuracy close to laboratory methods and released predicted structures for over 200 million proteins, and its developers shared the 2024 Nobel Prize in Chemistry. Predicted structures still need experimental checks, but they now guide research at a pace experiments alone could not match.
Personalised medicine and pharmacogenomics
Sequencing a patient's tumour can reveal which mutation drives it, and targeted drugs exist for several such mutations, so two patients with the same cancer type can receive different treatments. Pharmacogenomics applies the same logic to drug response: genetic variants change how quickly people process medicines. The enzyme CYP2D6, for instance, converts the painkiller codeine into morphine, its active form. People with low CYP2D6 activity get little relief because the conversion barely happens, while ultra-fast converters can build morphine up so quickly that normal doses become risky. Knowing a patient's genotype lets a prescriber pick a different drug or dose.
Genomics outside medicine
Agriculture links traits to genes so breeders can select drought tolerance or disease resistance directly, rather than waiting seasons for traits to appear. Biofuel research hunts microbial genomes for enzymes that convert plant waste into fuel, and environmental scientists track how pollution reshapes the microbial communities of soil and water. Forensics uses genomics too: after the 2001 anthrax letter attacks in the United States, investigators compared the genome of the mailed strain with laboratory samples to identify its source. Mitochondrial DNA, which passes down the maternal line almost unchanged, is widely used to trace ancestry and identify remains.
Ethics, law, and fairness
Genetic data are unusually personal: they describe not only you but your relatives, who never consented. Concerns include privacy, discrimination by employers or insurers, and ownership of samples. In the United States, the Genetic Information Nondiscrimination Act of 2008 bars health insurers and employers from using genetic test results, though it does not cover life or disability insurance. Fairness is a scientific problem as much as a moral one: most participants in genomic studies have been of European ancestry, so risk scores trained on that data predict less accurately for everyone else, one reason studies are being diversified.
Common mistakes to avoid
A GWAS hit is an association, not a diagnosis: most people carrying a risk variant never develop the disease. A proteome is not a readout of the transcriptome: splicing, translation control and chemical modification all break the equivalence between mRNA amounts and protein amounts. A screening test's headline accuracy hides its false-positive rate, which matters more the rarer the disease. And predicted protein structures are models: excellent starting points, but experimental structures remain the standard of truth. Treat every genomic claim as a probability, measured in a particular population.
Key takeaways
Genome data predict single-gene diseases well, but those cause only about 5% of the disease burden; common diseases are polygenic and environmental, so risk estimates are probabilities. GWAS finds variants associated with disease and points research at genes worth studying. Transcriptomics (RNA-seq, microarrays, single-cell) measures expression; proteomics (mass spectrometry) measures the proteins doing the work, and the two diverge through splicing and modification. Protein signatures can screen for disease once false positives and rarity are accounted for. Applications span targeted therapy, pharmacogenomics, breeding, biofuels and forensics, all governed by privacy and fairness concerns.
Flashcards
Flashcard
Scans many genomes to find variants linked to a disease
Test your understanding
Question 1 of 5
Why can most common diseases not be predicted from a single gene test?
Question 2 of 5
A GWAS finds a variant in 32% of patients and 24% of controls. What does this show?
Question 3 of 5
Why can protein levels in a cell differ from its mRNA levels?
Question 4 of 5
A screening test flags 88 of 100 disease samples but also 6 of 100 healthy samples. What are its sensitivity and false-positive rate?
Question 5 of 5
Why are genetic risk scores often less accurate for people of non-European ancestry?
