Genome analysis, genetic screening and gene therapy

Sequencing and annotation

The genome is the complete set of DNA in a cell, coding and non-coding alike. Sequencing determines the order of its bases; annotation, a separate process, identifies genes and regulatory regions within that sequence.

The original method, chain termination sequencing, copies a template in the presence of a few modified nucleotides that stop the polymerase wherever one is incorporated. That gives fragments of every possible length, each ending in a known base; separate them by length, read the terminating base of each in turn, and the sequence falls out. Modern high-throughput sequencing reads millions of fragments in parallel: the commonest platforms add one fluorescently labelled nucleotide at a time, others read long single molecules as they pass through a nanopore, and the reads are assembled by computer.

Costs have fallen by several orders of magnitude. The Human Genome Project ran from 1990 to 2003 and cost in the region of two billion pounds to produce one composite human sequence. A human genome can now be sequenced in about a day for a few hundred pounds, which is what made routine clinical and forensic use practicable.

Genome sequence allows researchers to predict coding regions, but annotation is still required. This is relatively direct in many bacteria. Viral genomes can be compact and may use overlapping genes, RNA processing or polyproteins, so sequence alone does not provide a simple list of mature proteins.

Predicting a eukaryotic proteome from a genome sequence is harder again. Introns must be identified and removed, alternative splicing allows one gene to give several polypeptides, large stretches are regulatory rather than coding, and post-translational modification alters the product further. The contrast between prokaryotic and eukaryotic genomes is a stated specification point on more than one board.

The uses are real enough. Sequencing pathogens identifies strains and tracks how an outbreak spreads, which is how variants of SARS-CoV-2 were followed almost in real time. Comparing genomes between species gives evidence about evolutionary relationships. Comparing them between individuals identifies the alleles associated with disease and, increasingly, which drug a particular patient will respond to. Public databases such as GenBank, UniProt and the Protein Data Bank hold these sequences and solved protein structures, so a new sequence can be compared with everything already known in minutes and a gene's likely function predicted from similar sequences.

One named application follows from that contrast. Where a bacterial genome can be annotated reliably, sequencing a pathogen produces a predicted list of the proteins it can make. Among those proteins are the surface ones an immune system would meet first, and a surface protein that every strain carries and that the host does not is a candidate antigen for a vaccine. Screening a list on a computer is faster and safer than culturing the organism to find out, which matters most for pathogens that are dangerous to grow or that will not grow in culture at all. The same route is far harder in a eukaryotic parasite, for the reasons given above for eukaryotic genomes.

Probes and microarrays

Once you know a sequence you can go looking for it, and the tool for that is a DNA probe: a short single strand of DNA, typically twenty to a few hundred bases, complementary to the sequence you are hunting, with a label attached. The label is either a radioactive isotope, detected by exposing photographic film, or a fluorescent tag, detected under ultraviolet light.

The method is hybridisation. Sample DNA is made single-stranded by heating, the labelled probe is added, and wherever the probe meets a complementary sequence it binds by hydrogen bonding between complementary bases. Unbound material is washed away, and the label shows where the probe stuck.

Describe the drawing

At the top, a short single-stranded DNA probe reading T A C G G T with a labelled tag drawn as a disc on one end, paired base by base against a longer target strand reading A T G C C A, with the hydrogen bonds drawn as short vertical ties between them. Below and to the left, a microarray is drawn as a grid of thirty-five spots, six of them lit and the rest dark, with a note that each spot carries a different known sequence and that a lit spot means the sample contained the complement of that sequence.

A DNA microarray is a small plate carrying thousands of different single-stranded probes fixed in a grid, each spot a known sequence at a known position. Labelled DNA from the sample is washed over it, unbound material is rinsed off, and the pattern of spots that fluoresce is read by a scanner. Because the position of each probe is known, the pattern is a direct read-out of which of those thousands of sequences the sample contains.

An array has two distinct applications, and they answer different questions. Washing DNA over it identifies which alleles the person carries, which is a genotype. Making cDNA from the mRNA of a tissue and washing that over it identifies which genes that tissue is transcribing, which is a pattern of expression. The second is how the expression profile of a tumour is compared with healthy tissue from the same patient, which in turn is how some cancers are now matched to a drug.

DNA probe

A short single-stranded length of DNA, labelled radioactively or fluorescently, with a base sequence complementary to a target sequence.

Hybridisation

The binding of a probe to a complementary single-stranded target sequence by hydrogen bonds between complementary bases.

DNA microarray

A plate carrying many different DNA probes at known positions, allowing thousands of sequences to be tested for at once.

Screening, counselling and what a result means

Genetic screening is testing an individual for particular alleles. In the United Kingdom every newborn is offered the heel-prick blood spot test in the first week of life, which as of 2025 screens for nine conditions including cystic fibrosis and sickle cell disease (the NHS reviews the list, so check the current one), because early treatment changes the outcome in every one of them. People may be screened as adult carriers before starting a family, as an unborn fetus by chorionic villus sampling or amniocentesis, as an embryo during IVF before implantation, or predictively for a condition that has not yet appeared.

Prenatal tests differ in timing and risk. Chorionic villus sampling takes a small sample of placenta, usually at 11 to 14 weeks, so the result comes early. Amniocentesis draws amniotic fluid containing fetal cells, usually from 15 weeks. Each carries a small risk of miscarriage, about one in two hundred or less. Pre-implantation genetic diagnosis tests one or a few cells removed from an embryo made by IVF, so that only an unaffected embryo is implanted. A positive prenatal result leaves parents deciding whether to continue the pregnancy, which is why counselling comes before the test as well as after it.

Genetic counselling is the conversation around the test. A counsellor explains what the test can and cannot show, works out the risk to the person and to any children, sets out the options, and supports whatever decision follows. The professional standard is that counselling is non-directive: the counsellor does not recommend a course of action, because the decision belongs to the family and the counsellor does not have to live with it.

What a positive result means differs greatly between conditions. Two contrasting examples follow.

A woman who tests positive for a pathogenic BRCA1 variant has a lifetime risk of breast cancer of roughly sixty to seventy per cent, against something like twelve to fifteen per cent for women in general. That is a large increase and it is not a diagnosis. Some women with the variant never develop cancer; most women who develop breast cancer do not carry it. The result changes what she might reasonably do, whether more frequent screening, risk-reducing surgery or nothing, and it does not tell her what will happen.

A person who tests positive for the Huntington's disease allele is in a different position. Penetrance is essentially complete, so the result does say that the disease will develop, though not when. The reason many people at risk choose not to be tested is precisely that: there is no treatment that changes the course of the disease, so the information alters the years before onset without altering the disease.

The ethical arguments are set out below on both sides rather than resolved. On one side, knowing allows planning, informed reproductive decisions and earlier surveillance, and withholding a test treats an adult as unable to handle information about their own body. On the other there is a recognised right not to know, testing one person discloses information about relatives who did not consent, and a predictive result can affect employment and insurance. In the UK, under the Code on Genetic Testing and Insurance as it stands in 2025, insurers may not require predictive results, with one agreed exception for Huntington's disease on life cover above half a million pounds; the code is voluntary and periodically reviewed. Prenatal screening carries a further argument, raised most sharply by disabled people's organisations: that screening for conditions such as Down's syndrome communicates a judgement about the worth of lives lived with them, whatever the intentions behind it.

Personalised medicine: matching the drug to the genotype

Conventional prescribing matches a drug to a diagnosis: patients with the same diagnosis receive the same molecule at a standard dose, and response and adverse effects are assessed afterwards. Personalised medicine, also called stratified medicine, uses the patient's genotype, or the genotype of their tumour, to select the drug and the dose before treatment begins. The branch dealing with drug handling is called pharmacogenomics. It uses sequencing, probes and microarrays.

The clinical literature increasingly uses 'stratified' rather than 'personalised', because no new molecule is designed for an individual patient. Patients who share a diagnosis are sorted into groups by genotype, and each group is treated differently. Three worked examples follow.

Which drug: HER2 and trastuzumab. Between one in six and one in five breast tumours carry extra copies of the HER2 gene and cover their surface with the growth factor receptor it codes for, so the pathway stimulating division is continuously active. Trastuzumab is a monoclonal antibody with a binding site complementary to that receptor. In a tumour that over-expresses HER2 it blocks the signal and marks the cells for destruction, and it substantially improved survival for that group of patients. In a tumour that does not over-express HER2, there is no target for the antibody, so the drug confers no benefit while still incurring cost and a risk of cardiac damage. The tumour is therefore tested before the drug is prescribed; a test whose purpose is to decide whether a particular drug should be used is called a companion diagnostic. The variant tested here is somatic, acquired by the tumour, rather than the germ-line genotype the patient inherited.

What dose: TPMT and azathioprine. Azathioprine and mercaptopurine, used in leukaemia, inflammatory bowel disease and after transplants, are broken down by the enzyme thiopurine methyltransferase. Roughly one person in three hundred inherits two low-activity alleles of the TPMT gene and cannot clear the drug; on a standard dose it accumulates until their bone marrow stops producing blood cells, which is a life-threatening reaction. Around one in ten carries a single low-activity allele and clears it slowly. Testing enzyme activity or genotype before the first prescription lets the deficient patient be given a small fraction of the usual dose, or a different drug, and the intermediate patient be started on a low dose and monitored. The drug is unchanged; the genotype determines the dose.

Where the case is weaker: warfarin. Warfarin, the anticoagulant, is handled by the enzyme CYP2C9 and acts on the enzyme VKORC1, and common variants of both genes shift the dose a patient needs by several times, which suggests a case for genotyping every patient before treatment. Trials comparing genotype-guided starting doses against the existing practice of starting at a low dose and adjusting on blood clotting tests have not shown a clear benefit, because the monitoring identifies the same patients within a week or two. The mechanism is well established and the clinical benefit is not: an improvement in outcome has to be measured rather than inferred from the mechanism.

The benefits are worth stating plainly, since boards ask for them as an argument. Fewer patients are given a drug that was never going to work for them, so their disease is not left untreated during a trial of it. Serious adverse reactions, which fill a substantial number of hospital beds, are avoided rather than treated. Expensive drugs go to the patients who will respond, which is the only way health systems can afford them. And a drug that fails in an unselected trial can be rescued if the subgroup it helps can be identified.

The limitations belong in the same answer. Sequencing and testing cost money and staff time, and a companion diagnostic has to be paid for on top of the drug, so a health service has to be shown the saving before it will fund the test. The reference data are unequal: the great majority of the genomes in the databases that variant effects are estimated from come from people of European ancestry, so a variant common in West Africa or East Asia is more likely to be reported as of unknown significance, and risk scores built on those databases predict less well for the patients who already get worse care. And most conditions are not single-gene at all. Heart disease, diabetes and depression are polygenic, with hundreds of variants each contributing a very small amount, on top of diet, exercise, poverty and age, so a genome shifts a probability rather than naming a treatment. Personalised medicine works best where a single gene decides a single thing, and that describes a minority of medicine.

Personalised medicine

Choosing which drug to give a patient, and at what dose, using their genotype or the genotype of their tumour, rather than treating everyone with the same diagnosis alike; also called stratified medicine.

Pharmacogenomics

The study of how a person's genotype affects the way they respond to a drug, including how quickly they break it down.

Companion diagnostic

A test carried out to decide whether a particular drug should be given to a particular patient, such as testing a tumour for HER2 before prescribing trastuzumab.

DNA profiling, and the statistic that goes with it

About 98 per cent of the human genome does not code for polypeptides, and scattered through it are short sequences repeated over and over, one after another. These are variable number tandem repeats (VNTRs). The sequence of the repeated unit is the same in everybody; the number of repeats at any given locus varies enormously between people and is inherited like any other allele, one from each parent.

That variation is the basis of DNA profiling. A sample is taken: blood, semen, saliva, a hair root, a few cells left on a surface. Its DNA is extracted and the chosen loci are amplified by PCR, using primers that bind to the constant sequences flanking each repeat region, so the length of each product depends only on how many repeats that person has. The products are separated by electrophoresis, and the pattern of lengths is the profile. The UK standard in use since 2014, called DNA-17, examines sixteen such loci plus a marker that indicates sex.

Describe the drawing

The upper half shows one locus in two people, drawn as a row of identical small blocks standing for a short sequence repeated over and over. Person A carries four repeats on one chromosome and seven on the other; person B carries five and nine, and the counts are printed at the right. The lower half shows the fragments run out on a gel: each person gives two bands, the longer allele stopping nearer the well and the shorter one running further, so the four bands sit at four different heights and the two people are told apart at a glance.

Because profiling uses non-coding repeats, a profile carries little information about the person: not their appearance, not their health, and not their ancestry in any detailed sense. That is a deliberate design choice, and it is part of why profiling is treated differently in law from whole-genome sequencing of suspects.

The match probability requires careful statement. A full profile at sixteen loci gives a probability of the order of one in a billion that an unrelated person chosen at random would match. That number is the probability of a matching profile given that the person is not the source. It is emphatically not the probability that the person is not the source given that the profile matches. Reversing those two is called the prosecutor's fallacy and has contributed to miscarriages of justice. State the conditional the correct way round when interpreting a match probability.

Three further limitations apply to any profile. Relatives share more alleles than strangers, so the one-in-a-billion figure does not apply to a brother. Degraded or very small samples give partial profiles at fewer loci, and the match probability weakens by orders of magnitude as loci drop out. And a match establishes that this person's DNA is in that sample; it says nothing about when it got there or how, which is a matter for other evidence entirely.

Forensic work and paternity are the uses everyone names, and they are not the only ones the specifications ask for. In medical diagnosis a profile can identify which variant of a pathogen an infection is, or reveal that a tumour sample carries a mutation the surrounding tissue does not, which changes the treatment. In animal and plant breeding it establishes pedigree and identifies which individuals carry a wanted allele, so a breeder selects on the genotype instead of waiting a generation to see the phenotype; it also lets a breeder avoid mating close relatives, which is how inbreeding is kept down in small populations. And across a population, comparing profiles measures genetic variability: many different alleles at the loci examined means a varied gene pool, and few means a narrow one, which is the number a conservation programme watches when a species is down to its last individuals.

Gene therapy, and why one kind is contested

Gene therapy means treating a disease by altering the genetic material of the patient's cells, usually by supplying a working allele where the patient has two faulty ones, sometimes by silencing a harmful allele or by correcting it directly. Getting the DNA into cells requires a vector: most often a modified virus, stripped of the genes that let it cause disease and carrying the therapeutic gene instead, or a liposome, a small sphere of phospholipid that fuses with the cell-surface membrane.

Describe the drawing

Two columns comparing the two kinds of gene therapy. The left column, somatic, describes delivering a working allele to the affected tissue only, and beneath it three discs stand for the patient, a child and a grandchild, with only the first filled in to show that only the patient is altered. The right column, germ line, describes delivering the allele to a gamete or a zygote before the cells have specialised, and all three discs are filled in to show the alteration passing down every generation.

Somatic gene therapy alters body cells. It has produced genuine treatments. Children with severe combined immunodeficiency have been treated by removing their bone marrow stem cells, inserting a working allele, and returning them; Luxturna, licensed in 2017, delivers a working RPE65 allele into the retina and restores useful vision in a form of inherited blindness; Casgevy, approved in the UK in 2023, uses gene editing on a patient's own blood stem cells to treat sickle cell disease and β-thalassaemia. Gene editing is genetic engineering that inserts, deletes or replaces DNA at a chosen site in the genome. In CRISPR-Cas9, a short guide RNA complementary to the target leads the Cas9 nuclease to cut both strands there, and the cell's repair of the cut either disrupts the sequence or copies in a supplied one. Cystic fibrosis remains the standard teaching example and the standard cautionary one: a working CFTR allele can be delivered to airway epithelium in a liposome or a virus, but epithelial cells are replaced continually, so the treatment has to be repeated and the clinical benefit so far has been small.

Somatic therapy has four limitations. Delivery is inefficient, the immune system may attack a viral vector, and a vector that inserts at random may land inside another gene: in an early trial for severe combined immunodeficiency several children later developed leukaemia because the vector had integrated beside a proto-oncogene. Cost raises a question of justice by itself, with single-dose treatments priced in the millions.

Germ line gene therapy alters a gamete, or a zygote, or an early embryo, so that every cell of the resulting person, including the cells that will make their gametes, carries the change. It is technically possible; in 2018 He Jiankui announced the birth of twin girls whose embryos he had edited, and was subsequently imprisoned in China for three years. In the United Kingdom, clinical use of heritable genome editing is not permitted. Laws and regulatory approaches differ between jurisdictions and can change, so name the jurisdiction and date when discussing legal status.

The case for is that some conditions cannot be treated any other way. A disease that has already damaged an organ during development cannot be undone by treating cells afterwards, and for a couple who both carry two copies of a recessive allele, no embryo selection is possible because every embryo will be affected. Correcting the allele once removes the disease from that family permanently rather than treating each generation in turn, which is arguably kinder and certainly cheaper.

The case against rests on four points that are usually made together. Consent cannot be obtained from the person who is altered, nor from any of their descendants, and they are the people who bear the risk. Editing is not perfectly precise, and an off-target change would be inherited along with the intended one, spreading a new problem through a family line. There is no agreed line between preventing disease and selecting for traits, and once the technique exists the pressure to move the line is unlikely to be zero. And because the technique would be expensive, an ability to give children biological advantages would follow wealth, which some argue would entrench inequality in a form no policy could later reverse.

A third position, held by several national academies, is that germ line editing is not wrong in principle but is not acceptable at present, because the safety data and the mechanisms of oversight do not yet exist. That is not a restatement of either side: 'not yet, under these stated conditions'.