Genetics and biotechnology, oriented

A gene is a defined stretch of DNA. Some genes carry instructions for a protein, while many others make no protein and help control other genes instead. Protein-coding genes are only a small fraction of the genome. Humans are diploid, with two chromosome sets; a grocery-store strawberry carries eight sets, 56 chromosomes in all.

A locus is a position on a chromosome; an allele is one sequence version found there. You carry two alleles at each locus on the numbered chromosomes, one from each parent, even when the population holds many more. Your genotype is what you carry, written as symbols such as Bb or as bases such as CT. Your phenotype is what shows, and it adds the environment.

Two identical alleles make you homozygous, two different ones heterozygous. In a heterozygote a recessive allele can stay completely hidden, because one working copy supplies enough function. That is what a carrier is: no sign of the condition, yet the allele can still pass to a child. Say two people both show a dominant trait: one may be BB and the other Bb, and how they look typically will not tell them apart.

The trap is reading "dominant" as common, strong or beneficial. Dominance arises when half the normal product is not enough or when the altered copy does something new and harmful, and neither route says how often the allele occurs. Geneticists mostly use "dominant disorder" for rare conditions.

Rule: treat dominance as a statement about two alleles inside one heterozygote, never about how often an allele occurs.

Each nucleotide holds a phosphate, a deoxyribose sugar and one of four bases. Sugar and phosphate alternate along the backbone, so the information sits entirely in the order of the bases. A pairs with T and G with C, each rung joining a purine to a pyrimidine, which is why one strand's sequence fixes the other's. RNA uses ribose instead, and uracil in place of thymine.

The backbone has a direction, 5-prime to 3-prime, and the two strands run opposite ways. DNA polymerases build only in the 5-prime to 3-prime direction and cannot start a strand from nothing, so primase first lays down a short RNA primer. The leading strand is made in one continuous piece, while the lagging strand is built in Okazaki fragments, with a new primer about every 100 to 200 nucleotides; the primers are later removed and DNA ligase seals the nicks. Each new double helix keeps one old strand and one new one. Accuracy comes in layers: the polymerase's choice of base, proofreading, then mismatch repair. At the very end of a linear chromosome the last gap cannot be filled, so telomeres shorten unless the enzyme telomerase adds repeats back.

The trap is writing a complement in the wrong direction. Say you are given 5'-GGATCA-3'. It pairs base for base with 3'-CCTAGT-5', so in the standard 5-prime to 3-prime order you write 5'-TGATCC-3'.

Rule: pair the bases, then reverse the string, because every sequence is written 5-prime to 3-prime.

About 150 base pairs wrap around each histone core to form a nucleosome, and further folding packs roughly six feet of DNA per human cell into a nucleus. Humans carry 22 numbered autosome pairs plus the sex chromosomes.

Cells spend most of their lives in interphase, and DNA is copied once, during S phase. The two copies, called sister chromatids, stay joined at the centromere. Mitosis pulls them apart into two daughter cells identical to each other and to the parent. Checkpoints can pause the cycle when DNA is damaged or a chromosome is not yet attached to the spindle. Cancer is partly a failure of that control: an oncogene can act like a stuck accelerator, and a switched-off tumor suppressor like a failed brake.

Meiosis runs two divisions after one round of copying. Meiosis I separates homologs, the maternal and paternal versions of each chromosome, and meiosis II separates sisters, so each egg or sperm ends up with one set. Crossing over between homologs makes chromosomes no parent carried, and each pair lines up independently of the others. When a pair or a set of sisters fails to separate, a gamete gains or loses a chromosome.

The trap is counting chromatids as chromosomes. Picture a cell that has just finished S phase. It holds twice the DNA, yet you still count each duplicated chromosome, two sisters sharing one centromere, as one until the sisters are pulled apart.

Rule: count centromeres, not DNA copies, and ask which division separates which pairs.

Transcription starts when RNA polymerase binds a promoter. The polymerase builds RNA complementary to the template strand, so the transcript matches the other strand, the coding strand, except that U replaces T. A eukaryotic message gets a methylated guanosine cap at its 5-prime end and a poly-A tail at its 3-prime end, and both help protect it. The spliceosome cuts out introns and joins exons, and joining different exon sets lets one gene yield several proteins, so counting genes underestimates what a genome can make. Bacteria have no nucleus, so a ribosome can start translating a message still being transcribed.

Ribosomes read three-base codons. Three bases give enough combinations for the 20 amino acids plus stop signals, so most amino acids have more than one codon. AUG marks the start and fixes the reading frame. A transfer RNA's anticodon pairs with each codon while it carries the amino acid an enzyme attached to it. Because the code is nearly universal, bacteria or yeast given a human gene can make the human protein, though variant codes have turned up in over 30 lineages, human mitochondria among them.

The trap is looking up codons from the template strand or from the anticodon. If the coding strand reads 5'-ATGTTCGGA-3', you write the message as 5'-AUGUUCGGA-3': copy the coding strand and swap T for U, with no complementing at all.

Rule: convert to messenger RNA letters first, then read codons 5-prime to 3-prime in the frame the start codon sets.

Bacteria group related genes into operons, read as one message, so one signal can switch a whole pathway, such as the genes that let E. coli take in and break down lactose. In eukaryotes, transcription factors bind regulatory DNA to raise or lower transcription. An enhancer can sit anywhere from about 100 bases to millions of bases from its gene, because the DNA bends to bring it to the promoter. Control continues later: microRNAs silence messages, and adding phosphate or a ubiquitin tag changes or removes finished proteins. Switching different gene sets on and off during development is what makes a brain cell differ from a liver cell.

Epigenetic marks change how DNA is used without changing its letters. Methyl groups on DNA can help switch genes off, chemical groups on histones loosen or tighten the packing, and marks kept through cell division help each cell type stay specialized. Imprinting silences one parent's copy of certain genes, and in female mammals one X is largely silenced in each cell.

The trap is treating a changed mark as an inherited one. Say you read that smoking, diet or pollutants shifted a person's marks; some such shifts do harm. That alone says nothing about their children: mammalian germ cells go through genome-wide reprogramming that removes DNA methylation, and only under some circumstances do marks carry into the next generation.

Rule: ask which signals a gene responds to, and at which step, and demand separate evidence before calling any mark inherited.

In Mendel's model each gamete carries one allele of each gene, the two kinds formed in equal numbers, so crossing two Aa parents gives 1 AA, 2 Aa and 1 aa: a 3 to 1 phenotype ratio when A is dominant. For two independent genes, you multiply their separate probabilities rather than filling in 16 boxes. Mendel's test cross to the recessive form reveals an unknown genotype, and a chi-square test asks whether counts stray further than chance allows, with degrees of freedom equal to the classes minus one when you estimate nothing from the data.

Departures have causes. Mendel's early defenders already stressed that a heterozygote can look intermediate while segregation still works, which gives phenotypes of 1 to 2 to 1. A genotype that dies before it is counted turns 1 to 2 to 1 into 2 to 1. Epistasis lets one gene mask another: in Labrador retrievers, a separate gene can stop pigment reaching the fur, giving yellow dogs whatever the color genotype. Many small-effect genes plus environment give a continuous range of values.

Genes close together on one chromosome are linked, so parental combinations outnumber recombinant ones. The recombinant share measures distance, 1 percent being one centimorgan. Early fruit-fly mapping showed it topping out near 50 percent, the value free assortment gives.

The trap is treating a skewed ratio as a failed experiment. Mendel's round, yellow hybrids gave 315, 101, 108 and 32 seeds, near 9 to 3 to 3 to 1 but not exact, and scatter that small is what chance gives. A large cross that keeps giving 2 to 1 instead of 3 to 1 is different: that gap is evidence about the genes, such as a lethal genotype.

Rule: predict with the product rule, test the deviation, then name the mechanism behind it.

The X chromosome carries far more genes than the Y, so a male has a single copy of most X genes, and one recessive allele is enough to show. Affected males therefore far outnumber affected females in X-linked recessive conditions, and a carrier mother usually shows no signs because her second X makes up for the altered one.

A pedigree draws a family as one chart, squares for males and circles for females, shaded when affected. An autosomal recessive condition can appear in the child of two unaffected carrier parents, and related parents raise that chance, because each can carry the same rare allele from a shared ancestor. An autosomal dominant condition usually shows in every generation, though a new variant can make someone the first affected person in a family, and reduced penetrance means some people with the variant never show signs. Each pregnancy is a fresh draw, so an earlier child, affected or not, changes nothing about the next.

The trap is choosing the mode that fits most of the chart. Say your chart shows an affected father with one unaffected daughter. Under X-linked dominant inheritance every daughter of an affected father inherits the condition, so with full penetrance that one daughter rules the mode out, however well the rest of the family fits it.

Rule: test every mode against every transmission, and keep only the modes no one in the chart contradicts.

A substitution swaps one base. Because several codons share an amino acid, it may change nothing, swap one amino acid, or create a stop codon that cuts the protein short. An insertion or deletion not divisible by three shifts the reading frame and garbles everything after it; one that adds or removes whole codons keeps the frame, though even a three-base deletion can destroy function. A germline variant can pass to children, while a somatic one stays in the body cell where it arose and that cell's descendants. Copying errors cause mutations even without exposure to a mutagen, and most variants have no effect on health.

In a population, count alleles, not people, since each person carries two copies. With allele frequencies p and q, the Hardy-Weinberg expectation is p squared, 2pq and q squared, provided there is no new mutation, no selection, no migration, random mating and a large population. Genetic drift changes frequencies by chance, fastest in small groups, and a population descended from a few founders can carry an allele that is rare elsewhere. Gene flow pulls populations together, and selection can even keep a harmful allele common when carriers of one copy outsurvive both homozygotes.

The trap is plugging a disease rate in as the allele frequency. Say you read that 1 in 90,000 births shows a recessive condition. That rate is q squared, so q is 1 in 300, and the carrier share, 2pq, comes to about 1 in 150.

Rule: for a recessive condition the affected rate is q squared, so take its square root before computing anything else.

The Human Genome Project ran from 1990 to 2003, and its 2003 sequence covered about 92 percent of the genome; the most repetitive stretches were finished only in 2022. The reference genome is an agreed standard assembled from multiple anonymous donors, not one person's DNA, and pangenomes now combine many genomes to capture more variation.

Sanger sequencing still serves well for short stretches. Next-generation instruments run huge numbers of reactions side by side, giving reads of about 150 to 300 bases that are hard to place in repetitive DNA. Long-read methods average over 10,000 bases per read and can span a repeat, at a higher per-read error rate. Coverage, or depth, is how many reads cover a position, and a base seen in many independent reads earns more confidence. A variant on one of the two copies is expected in about half the reads, on both copies in nearly all of them. Reads are mapped onto a reference to find differences, or assembled from scratch when no reference exists.

The trap is reading a negative result beyond its reach. Say you see a clean exome. Exome sequencing reads the roughly 1 percent of the genome made of exons and catches most known disease-causing changes, yet it cannot see a variant outside the exons that alters gene activity, so all you know is that nothing was found in what was read.

Rule: before trusting a call or a clean result, ask what was read, how deeply, and against which reference.

Each cycle of the polymerase chain reaction (PCR) heats DNA to separate the strands, cools it so two primers can bind on either side of the target, and lets a polymerase extend them. Every new strand serves as a template in the next cycle, so 30 to 40 cycles turn a few molecules into more than a billion copies. A heat-stable polymerase from the bacterium Thermus aquaticus is what made machine-run cycling practical, and primers of about 18 to 24 bases keep the product specific.

DNA is negatively charged along its backbone, so in gel electrophoresis an electric field pulls fragments through a gel that acts as a sieve. Small fragments travel furthest, and a size marker run alongside turns distance into length. A labeled probe then pairs with one target sequence among many.

For cloning, you cut DNA with restriction enzymes at short recognition sites, often leaving sticky overhangs that pair with any piece cut by the same enzyme, and DNA ligase seals the joined backbone. A plasmid carries an origin of replication, a site for the insert and a selection marker, so only cells that took it up grow on the antibiotic. A human gene goes into bacteria as DNA copied from its messenger RNA, because bacteria cannot splice out introns.

The trap is trusting a band without checking the blank. Say your sample lane shows a strong band. PCR copies any matching molecule, so a trace of stray DNA gives a product too. A no-template control, run with no sample at all, must stay empty, and a band there means contamination.

Rule: check the no-template control before you read a single sample lane.

A genome-wide association study compares people who have a condition with similar people who do not, across many single-letter variants. Testing about a million variants at the usual 0.05 cutoff would flag tens of thousands by chance, so studies require p below 5 times 10 to the minus 8. A flagged marker often only travels with the real causal change inside a block inherited together, and such variants typically raise disease odds by about 10 percent or less. A polygenic risk score adds many small effects into one number that places a person relative to others, but it sorts individuals poorly and works less well across ancestries.

Genetic data raises its own questions. Relatives share DNA, so one person's results say something about family members, and a genomic sample can never be truly anonymized. The US Genetic Information Nondiscrimination Act bars health insurers and employers from using genetic information in coverage and job decisions, but it does not apply to employers with fewer than 15 employees or to life, disability or long-term care insurance. A forensic DNA match says how rare a profile is, not how likely someone is to be guilty.

The trap is reading heritability as a personal share. Say you read that a trait has a heritability of 0.7. That describes variation across one population in one environment; it does not mean the trait is 70 percent caused by genes in any one person. Hair color is highly heritable, and it is still easy to dye.

Rule: keep population statistics at population scale, and treat an association as an address to investigate, not a cause.

Bacteria store snippets of viral DNA as clustered regularly interspaced short palindromic repeats (CRISPR) and use RNA copies of them to steer an enzyme such as Cas9 against a returning virus. In the lab, you give a guide RNA about 20 bases matching the target, so retargeting means writing a new guide rather than designing a new protein. Cas9 also needs a short motif beside the target: the widely used version from Streptococcus pyogenes requires NGG, where N is any base.

Cutting is only half the edit. Non-homologous end joining rejoins the ends with small random insertions or deletions that often knock a gene out, while homology-directed repair copies a supplied template to write in a precise change but works mainly in dividing cells. Editing a patient's body cells affects one person; editing an embryo would reach descendants, and the National Institutes of Health does not fund gene editing in human embryos.

Gene cloning copies a DNA fragment in a vector, while reproductive cloning moves a body-cell nucleus into an egg whose own nucleus was removed; Dolly the sheep took 276 attempts. A pluripotent stem cell can become any adult cell type and a multipotent one only those of its own tissue, and adult cells can be reprogrammed into a pluripotent state.

The trap is assuming a perfect guide guarantees a cut. Say your guide matches its target perfectly. If no NGG sits beside that target, this Cas9 will not cut there, though engineered variants relax the requirement.

Rule: check the motif first, then choose the repair route that produces the change you want.