Genetics & Biotech curriculum

Some genes carry instructions for a protein, but many do not and help control other genes instead, and cells also make working RNA molecules, such as transfer and ribosomal RNA, that never become protein. The roughly 20,000 human protein-coding genes make up only about 1 to 2 percent of the genome, so a gene is a small, defined part of DNA, not DNA in general.

Your genome is your complete set of DNA: in humans, 23 chromosome pairs in the nucleus, while mitochondria hold a far smaller circular chromosome of their own. Genes are only part of that set, and protein-coding genes only about 1 to 2 percent of it, so reading a genome means far more than reading its genes.

Because you inherit one copy from each parent, you normally carry two alleles at each locus on the numbered chromosomes, even when a population holds many more. An allele can differ by a single base or by a longer segment, so two people can share a locus and still differ at one letter of it.

A genotype can be written as symbols such as Bb or as the actual bases, such as CT. Phenotype adds the environment, so one genotype can look different depending on diet or exercise. The link fails in reverse too: for a dominant trait, BB and Bb typically look the same.

A heterozygote can hide a recessive allele completely, so its phenotype cannot tell you whether that allele is there. A carrier of a recessive condition is exactly this case: one working copy supplies enough function for the cell, yet the hidden allele can still pass to the next generation.

Dominance arises in two ways: half the normal gene product may not be enough, or the altered copy may do something new and damaging. Neither route says how often the allele occurs; geneticists mostly use 'dominant disorder' for rare conditions, so a dominant allele can be uncommon.

Human body cells are diploid and eggs and sperm are haploid, but other arrangements are common. Male bees, wasps and ants develop from unfertilized eggs and are haploid, while some organisms are polyploid: a grocery-store strawberry is octoploid, with eight sets of seven chromosomes, 56 in all.

Genetics follows single genes and how traits or conditions pass between generations. Genomics studies all of a person's genes together, including how they interact with each other and with the environment, which is why complex diseases are mainly genomic questions. The suffix '-ome' comes from Greek for all, every or complete.

Along a strand the sugar and phosphate simply alternate, so the information lives entirely in the order of the four bases. RNA nucleotides differ in two parts: ribose replaces deoxyribose, and uracil takes the place of thymine.

That chemical direction is why each strand has a 5-prime end and a 3-prime end, and why the two strands of a helix run opposite ways. DNA polymerases can build only in the 5-prime to 3-prime direction, so the backbone's direction, not the bases, decides how each strand is copied.

Pairing is so specific that one strand's sequence fixes the other's, which is why a DNA sequence is written as a single string of letters. In RNA, adenine pairs with uracil instead. GC-rich stretches hold together more firmly, but melting measurements suggest the stacking of flat bases on one another, not hydrogen bonds alone, supplies most of that stability.

Because polymerases add nucleotides only in the 5-prime to 3-prime direction, opposite strand directions mean only one strand at a fork can be copied smoothly. The other is made in short Okazaki fragments, each needing a fresh primer about every 100 to 200 nucleotides.

The model's power was explanatory, not just descriptive. Pairing A with T and C with G across two strands meant one strand's sequence fixed the other's, so each strand could act as a template for building a new partner. That showed at once how DNA could be copied and serve as the cell's information molecule.

Adenine and guanine are purines; cytosine, thymine and uracil are pyrimidines. Since A pairs with T and C with G, every rung of the ladder joins one purine to one pyrimidine. Uracil is the pyrimidine RNA uses in place of thymine.

RNA swaps in ribose and uracil, is usually single-stranded, and works in several forms: messenger, transfer and ribosomal. Some viruses even use RNA as their genome. RNA is also broken down easily, so old messages do not linger and a cell can change which genes it expresses.

In PCR, heat separates the sample into single strands, an enzyme copies each one, and the cycle repeats 30 or 40 times to reach more than a billion copies. Hybridization turns the same pairing into a test: short designed DNA pieces bind only where the bases match, revealing whether a sequence is present.

About 150 base pairs wrap around a histone core to form each nucleosome, and nucleosomes then fold on themselves again and again. Unwound end to end, the DNA in one human cell would stretch about six feet. Packing is reversible: chromatin unwinds for replication and transcription and condenses most tightly for division.

Chromatin is not packed evenly. Regions being transcribed sit in a more open structure, while heterochromatin is compact, including the highly repetitive satellite DNA around centromeres. Histone marks track the difference: losing acetylation compacts chromatin, and methylation at certain histone sites goes with heterochromatin, which tends to be copied late in S phase.

The centromere is usually off-center, and its position gives each chromosome its shape and splits it into a short p arm and a long q arm. A telomere is one short sequence repeated over and over; cells that divide often, such as white blood cells, use an enzyme to keep theirs from wearing away.

Stained chromosomes show light and dark bands, and band numbers rise with distance from the centromere. A karyotype shows changes in chromosome number or structure that are visible under a light microscope; newer methods screen without a microscope by comparing a sample's DNA with a reference sample.

The 22 autosome pairs are numbered by size, though chromosome 21 turned out slightly smaller than 22. Current estimates put 900 to 1,400 protein-coding genes on the X and only 70 to 200 on the Y, so a male has a single copy of most X genes, which is the root of sex-linked patterns.

A trisomic cell has 47 chromosomes and a monosomic one 45. Most chromosome abnormalities arise as an accident in the egg or sperm, so every cell carries them; when the error happens after conception, only some cells are affected, a state called mosaicism.

A rearrangement is balanced when no genetic material is gained or lost, as in a reciprocal translocation where two chromosomes swap segments, and it may have no effect on health. Such changes can still be inherited, which is why parents are often tested when a child is found to carry one.

Only about 1 to 2 percent of human DNA codes for protein, and much of the rest has jobs. Sequences outside that share help regulate when, how and how much protein is made; a repetitive kind called alpha satellite makes up the centromere; and a short repeat caps each telomere. Pseudogenes, gene copies that lost their coding ability, can complicate genetic testing.

Cells spend most of their lives in interphase. Replication origins are first recognized in G1, and DNA is then copied once and only once per cycle during S phase, so a cell that reaches division carries two complete copies of every chromosome.

In prophase chromosomes condense and the spindle forms; the nuclear membrane breaks apart in prometaphase; copied chromosomes line up in metaphase; anaphase pulls them into two identical groups; and telophase rebuilds nuclear membranes. Cytokinesis, the split of the cell itself, comes last.

Replication doubles the DNA without doubling the chromosome count, because the two copies stay joined at the centromere as one replicated chromosome. Only when spindle fibers pull them to opposite sides do they become two separate chromosomes, one for each daughter cell, with identical DNA.

Chromosome separation works much the same in plants and animals, but the physical split differs sharply. A plant cell builds a cell plate between the two daughter nuclei in late anaphase, and the plate grows outward until it divides the parent cell, using membrane thought to come from Golgi-derived vesicles.

Checkpoints let the cycle pause when something is wrong, such as incomplete DNA synthesis or damaged DNA. The spindle assembly checkpoint holds a cell in metaphase until every chromosome is attached to the spindle. A cell with damage it cannot repair may instead destroy itself through apoptosis.

A proto-oncogene normally regulates division; mutated into an oncogene it can act like a stuck accelerator. A tumor suppressor works like a brake, so the danger comes when it is switched off. Most mutations in cancer genes never lead to disease, largely because DNA repair removes them as they occur.

Mitosis builds and replaces body cells, so its two daughters match each other and the parent. Meiosis runs the division steps twice, halving 46 chromosomes to 23 and shuffling genes along the way, which is why the eggs or sperm one person makes differ genetically from one another.

The first division is reductional: homologs separate while each pair of sister centromeres stays together. The second is equational, splitting the sisters much as mitosis does. Crossovers hold each homolog pair together until anaphase I, which is what lets the first division sort them correctly.

In the long prophase I, a protein structure called the synaptonemal complex zips each homologous pair together along its full length, forming a bivalent. It supports the exchange of material between homologs and the crossovers needed for correct segregation, and errors here can lead to miscarriage or infertility.

Crossovers do more than mix alleles. The physical links they leave, called chiasmata, hold each homolog pair together until anaphase I, and each pair needs at least one crossover to separate correctly. Faulty control of crossovers is a primary contributor to chromosome missegregation.

At the first meiotic division, the mother's and father's copies of each chromosome pair line up and separate independently of every other pair, so each gamete gets its own mix of maternal and paternal chromosomes. The shuffle works between chromosomes; the nearer two genes lie on the same chromosome, the more likely they are to travel together.

Each egg and sperm carries one haploid set already shuffled by meiosis, and fertilization joins two such sets into a diploid zygote. Mendel modeled this as egg and pollen cells of each type taking part in fertilization equally on average, so the two parents' combinations multiply rather than add.

Nondisjunction leaves gametes with an extra or a missing chromosome. In meiosis I a homolog pair fails to part; in meiosis II sister chromatids do, which the cohesion joining sisters normally prevents until anaphase II. Missing crossovers make the first kind of error more likely.

Egg formation divides unequally, so one large oocyte keeps the stored material the embryo will need while the small polar bodies are set aside. Error profiles differ too: the chance of a chromosome-number error rises with the mother's age, while a father's age is not thought to add that particular risk, because sperm are made anew throughout life.

A bacterium that copies its DNA and divides thousands of times leaves clones that all share the same DNA. Sexual reproduction instead joins haploid gametes shaped by crossing over, so offspring carry new combinations, and that variation helps a species respond to changing environments.

Each parental strand serves as the template for a new partner, so both daughter molecules hold one old strand and one new strand. PCR depends on the same rule, since every cycle builds new strands on the old ones and so doubles the target segment.

Depending on genome size, eukaryotic cells use several hundred to thousands of origins so that large linear chromosomes copy in a manageable time. Origins are licensed in G1 and fired in S phase, and only some fire; the rest stay in reserve in case replication forks stall.

The helicase opens the duplex in both directions from an origin, and separating the strands overwinds the DNA ahead into positive supercoils. If topoisomerases do not relax them the fork stalls, and behind the fork they also unlink the two daughter molecules so the genome can be divided.

Primase lays down a short RNA primer, which a DNA polymerase then extends. The leading strand needs only one primer per replication machine, but the lagging strand needs a new start about every 100 to 200 nucleotides, and those primers must later be removed and replaced.

The leading strand is made in one continuous piece, while the lagging strand is built in Okazaki fragments that each start near the fork and extend away from it. At a chromosome's end the last primer's removal leaves sequence that cannot be filled, so telomeres shorten each round.

As a polymerase finishes one Okazaki fragment, it runs into the RNA primer of the next and lifts it off as a flap. A flap endonuclease cuts the flap away, leaving a nick that DNA ligase seals, so separate fragments become one continuous strand.

Accuracy comes in three layers. The polymerase prefers the correct nucleotide, proofreading snips a mistake back off the growing 3-prime end, and mismatch repair fixes errors that slip past proofreading; in eukaryotes it relies on breaks in the new strand to tell it from the old one.

Because DNA polymerases cannot finish copying the ends of linear chromosomes, dividing somatic cells lose about 200 nucleotides per cycle. Telomerase carries its own RNA template to add repeats back; most somatic cells keep it off, while stem cells, germ cells and most tumors use it.

In bacteria, the core RNA polymerase must first pick up a sigma factor, forming the holoenzyme, before it can recognize and bind promoter DNA. Eukaryotes split the work among three RNA polymerases, and for a protein-coding gene the promoter sits upstream, where polymerase and transcription factors bind.

The polymerase builds RNA complementary to the template strand, so the transcript ends up matching the other strand, the coding strand, base for base except that U stands in for T. To predict a message, copy the coding strand and swap T for U rather than complementing the template twice.

As the polymerase moves, a short RNA-DNA hybrid about nine nucleotides long sits inside the transcription bubble while the transcript's 5-prime end peels away. In bacteria, transcription stops at an intrinsic terminator, a stable RNA hairpin followed by a U-rich stretch, or when the Rho protein reaches the polymerase.

The cap is a methylated guanosine added to the 5-prime end while transcription is still underway. It blocks enzymes that would chew the message from that end and recruits factors for processing, export and translation, so removing it opens the message to breakdown.

Making the tail takes two steps, cleavage of the new transcript and then polyadenylation, and in mammals the tail can reach 200 to 250 adenines. Binding proteins shield it from degradation, and shortening the tail, called deadenylation, can start the message's decay.

Introns are copied into the first transcript and then cut out, so only exons remain in the mature message. The spliceosome, involving at least 170 proteins, recognizes each intron by a GU at its start, a branch-point adenosine, and an AG at its end. In many genes the introns are far longer than the exons.

Joining a gene's exons in different combinations yields related but distinct messages, and each can be translated into its own protein. That is how about 20,000 protein-coding genes supply a far larger set of proteins, and why counting genes underestimates what a genome can make.

With no nuclear barrier, a bacterial ribosome can start translating a message while it is still being transcribed, and one message often carries several genes of an operon. A eukaryotic message is capped, spliced and given a poly-A tail in the nucleus, and only the finished message is exported to the cytoplasm for translation.

Three-base codons give 64 combinations: 61 specify amino acids and 3 are stop signals. Because those 61 codons serve only 20 amino acids, most amino acids have more than one codon, which is where the code's redundancy comes from.

The table is written in RNA letters, A, U, C and G, so any sequence containing T is still DNA and must be converted before you look it up. Read the codons in order, three bases at a time, and keep the tRNA anticodon separate in your head: it is the complement of the codon, not the codon itself.

Redundancy is organized around the codon's last base, which pairs with the tRNA's wobble position under looser rules. A modified wobble base such as inosine can read codons ending in A, C or U, so a single tRNA can serve several codons; in mitochondria one tRNA can read all four endings of a codon block.

The start codon AUG is read by a special initiator tRNA, which in bacteria carries formylmethionine. A stop codon codes for no amino acid, so the ribosome halts and the open reading frame ends there. These meanings are not perfectly fixed: some lineages have reassigned a stop codon to an amino acid.

Insertions or deletions in multiples of three keep the frame, but any other number shifts every codon downstream, which often creates an early stop codon and cuts the protein short. Because DNA has two strands, any stretch can be read in six frames, three in each direction.

Faithful decoding depends on two matches. An aminoacyl-tRNA synthetase first links each tRNA to the amino acid that fits its anticodon, and the anticodon then pairs with the codon in the ribosome. Both steps have to be right for the code to be read faithfully.

A charged tRNA enters the A site; if its anticodon pairs properly with the codon, a peptide bond forms with the chain held by the tRNA in the P site. The ribosome then shifts one codon, moving that tRNA to the P site and the emptied tRNA to the E site.

The standard code is used nearly everywhere, which is why bacteria or yeast that copy an inserted gene can make proteins used to treat human disease. It is not completely frozen, though: alternative codes have turned up in over 30 lineages of bacteria, eukaryotes and mitochondria, human mitochondria among them.

The lac operon is the classic case: genes for bringing lactose into the cell and breaking it down are expressed together. Sharing one message has a cost, too: a premature stop codon in an early gene can shut off the genes after it, an effect called transcription polarity.

Its three genes, lacZ, lacY and lacA, let E. coli take in and use lactose; lacZ makes beta-galactosidase, which splits lactose into glucose and galactose. The operon is switched on when lactose is present and glucose is low, which is why glucose, when available, wins.

These genes make enzymes for tryptophan, so the cell has no need for them when tryptophan is plentiful. Besides the repressor, bacteria use attenuation: with enough tryptophan, a ribosome translating a short leader passes its two Trp codons without pausing, and the RNA folds into a hairpin that ends transcription early.

Signals from the environment or from other cells activate transcription factors, which bind regulatory regions of a gene and raise or lower transcription. In bacteria activators and repressors are usually separate proteins, but in eukaryotes one DNA-binding protein can act as either, depending on its partners.

An enhancer can sit upstream or downstream of its target gene, even inside another gene, anywhere from about 100 bases to millions of bases away. The DNA bends to bring it close to the promoter, which is why a disease-causing change can lie well outside a gene's coding region.

A message that exists can still be blocked. MicroRNAs, short noncoding RNAs, bind the untranslated regions of a message and silence it, and RNA interference uses small double-stranded RNA pieces to stop a message being translated once its protein is no longer needed.

After a protein is made, enzymes can add or remove chemical groups, such as phosphate in phosphorylation, changing its stability, location in the cell or partners. To remove a protein, the cell tags it with ubiquitin and the proteasome breaks it down.

Each cell turns on only a fraction of its genes. Switching different sets on and off during development makes a brain cell differ from a liver or muscle cell, and epigenetic marks passed through cell division help each cell type stay specialized; muscle cells, for example, do not make bone-growth proteins.

Proteins attach methyl groups at specific spots on the bases, and the tags work by changing how DNA interacts with other proteins rather than by changing the letters. Because they can be kept as cells divide, they let a cell remember which genes are on or off. Bacteria use methylation differently, to tell their own DNA from invading DNA.

Chemical groups such as methyl or acetyl groups are added to or removed from histones, changing how tightly DNA wraps around them. Other proteins read these tags to decide whether a region is used, and particular marks track active genes, enhancers, silenced genes or heterochromatin.

The parent-of-origin mark is set by methylation while eggs and sperm form. Only a small share of human genes is imprinted, and they cluster in particular chromosome regions. Inheriting both copies of such a region from one parent can leave a person with no active copy, which can disrupt development.

Silencing one X balances gene dosage between XX females and XY males. In most body cells the choice is random, so a female is a patchwork of cells using different X chromosomes; in mice, though, the paternal X is silenced first, and only tissues outside the embryo itself keep that pattern. The patchwork helps explain why relatives with the same X-linked variant can differ in severity.

Smoking, diet, infection and pollutants can shift epigenetic marks, and some of those changes can do harm. The same flexibility is normal: marks are rearranged as cells specialize and can change across a lifetime, and that capacity to adjust appears to be required for health. None of this alters the DNA sequence itself.

Much of the epigenome is reset when parents pass on their genomes; mammalian germ cells go through genome-wide reprogramming that removes DNA methylation. Only under some circumstances do tags on egg and sperm DNA or histones carry into the next generation, which is why such claims need careful testing.

Mendel set two requirements for a test plant: constant, clearly different characters, and hybrids that could be shielded from stray pollen. Peas fit because the flower's keel protects its fertilization organs, and seed traits such as round versus angular could be counted in the thousands.

A hybrid's gametes are pure: each carries A or a, never both, and the two kinds form in equal numbers. Self-fertilizing the hybrid gives A + 2Aa + a, which looks like 3 dominant to 1 recessive because Aa resembles A. Mendel's pooled counts came to 2.98 to 1.

Mendel showed that how one pair of characters sorts does not depend on another, so a hybrid forms as many kinds of egg and pollen cells as there are possible constant combinations. The rule fits genes on different chromosomes; genes close together on one chromosome tend to travel together.

Crossing two Aa parents gives 1 AA, 2 Aa and 1 aa, a 1 to 2 to 1 genotype ratio but a 3 to 1 phenotype ratio when A is dominant. Real counts wander, especially with few offspring: one of Mendel's single plants gave 43 round seeds to 2 angular, and another 14 to 15.

Mendel's round, yellow hybrids gave 315 round yellow, 101 angular yellow, 108 round green and 32 angular green seeds, close to 9 to 3 to 3 to 1. Instead of filling 16 boxes, multiply: three quarters round times three quarters yellow gives nine sixteenths round and yellow.

Mendel used this logic to test his own model: he pollinated his round, yellow hybrids with the recessive angular, green form, so each offspring showed which kind of egg cell it came from. If the hybrid made its four egg types equally, the four offspring classes had to appear in equal numbers.

For independent events, multiply. Each child of the same two parents is a fresh, independent draw, so earlier children never change the odds for the next one. For outcomes that cannot occur together, add: the chance of showing a dominant trait is the chance of two dominant alleles plus the chance of one.

An F2 plant showing the dominant form can be AA or Aa, and only its offspring reveal which. When Mendel grew seeds from 519 yellow-seeded plants, 166 bred true and 353 split, about 2.13 to 1 against an expected 2 to 1. He also warned that egg cells that fail and seeds that never germinate push counts off target.

For each class, subtract the expected count from the observed count, square it, divide by the expected count, and add across classes. When no parameters are estimated from the data, the degrees of freedom are the number of classes minus one, and each expected count should be at least 5 for the test to be valid.

An intermediate heterozygote does not mean the alleles blended away: segregation still works, so two intermediates crossed together produce both parental types again. Mendel's early defenders stressed that his principle has nothing to do with whether the hybrid looks blended, as with a crimson and a white primrose giving a washy magenta.

Each allele makes its own slightly different product, and both show. In the ABO system, the A and B alleles encode enzymes that add different sugars to the same starting molecule, so a person carrying both builds both versions on their red cells rather than a blend of the two.

A locus can hold far more than two alleles across a population. Early fruit-fly work found a four-member series for one eye-color gene, and numerous alleles have since been identified at the human ABO locus. Any one diploid individual still carries at most two of them.

The same loss can skew a sex ratio. In fruit flies, a recessive lethal on the X kills every son that inherits it, because a male has no second X to cover it; mothers carrying such a lethal produce only half as many sons as daughters.

Mitochondrial disorders show the pattern clearly: a single change in mitochondrial DNA can affect several organ systems at once, with the strongest effects in the organs that use the most energy, such as the heart, brain and muscles.

Labrador retrievers show it: their coat-color genes code only for black or chocolate, yet recessive genes at a separate locus can stop pigment reaching the fur, producing yellow dogs whatever the color genotype. The second gene decides whether the first is seen at all.

When many genes each add a little, and the environment adds more, a trait spreads into a continuous range centered on an average instead of falling into a few classes, so it shows none of the simple single-gene ratios. Common conditions such as heart disease and diabetes involve many genes in this way.

Reduced penetrance means some people with a variant never show its signs; variable expressivity means those who do can range from mild to severe, even with a change in the same gene. Both probably reflect other genes, environment and lifestyle, mostly unidentified, which is why they make family histories hard to read.

Only egg cells keep their mitochondria at fertilization, so the line of mitochondrial DNA runs through mothers and can trace maternal ancestry far back. Mitochondrial DNA also repairs errors poorly, so variants that arise during life build up in body cells without being passed on.

In birds the female is the mixed sex, ZW, and males are ZZ, so the mother's egg decides the chick's sex. In bees, wasps and ants, males hatch from unfertilized eggs and are haploid. Where temperature sets sex, warming can tip a population toward one sex.

A female needs two altered copies to be affected, which is unlikely, while a male needs only one, so affected males far outnumber affected females. Fathers cannot pass X-linked traits to their sons, because a son inherits his father's Y chromosome.

A carrier mother has one altered X and one typical X, so each son has an even chance of receiving the altered copy, and each daughter an even chance of becoming a carrier. A carrier usually shows no signs, because her second, typical X can make up for the altered one.

One altered copy is enough in either sex, though females may have milder signs because they carry a second X. In some conditions affected males do not survive at all. As with every X-linked pattern, fathers never pass the trait to their sons.

The Y chromosome, almost 2 percent of a cell's DNA, likely holds only 70 to 200 protein-coding genes, and those tend to concern male sex determination and development. Because only males carry a Y, a Y-linked variant passes only from father to son; Y chromosome infertility is one known example.

A pedigree draws a family as one connected chart: squares for males, circles for females, a shaded symbol for an affected person and an open one for an unaffected relative. A double line between two partners marks a related couple, and the proband is usually the first family member to bring the concern to medical attention.

The parents of an affected child are usually carriers: each has one altered copy and no signs, which is why the condition often skips generations. Two copies are needed, one from each parent, so two carriers can have an affected child in any pregnancy, and the chance is the same each time. The gene sits on a numbered chromosome, not a sex chromosome, which separates this pattern from X-linked recessive inheritance.

One altered copy is enough, so each child of an affected parent has a 50 percent chance of inheriting it, and the condition usually shows in every generation. Two things break the picture: a new variant can make someone the first affected person in the family, and reduced penetrance means some people with the variant never develop signs, which makes a family history harder to read.

Test each mode against the whole chart and drop any mode that even one transmission contradicts, however well the rest fits. Look first for the telling cases. For example, under X-linked dominant inheritance every daughter of an affected father is affected, so a single unaffected daughter of an affected father rules that mode out.

Partners who share an ancestor can each carry a copy of the same recessive variant inherited from that ancestor, so their children face a higher chance of an autosomal recessive condition. Consanguinity does not change how common an allele is in the population; it raises the odds that two carriers of the same allele have children together, and the effect is largest for rare alleles.

Each pregnancy is a fresh draw: the chances are set by the parents' genotypes and start over every time, so an earlier child, affected or not, does not protect or doom the next one. Work out what each parent can pass on first, then combine the two for each child, and remember that an unaffected child can still be a carrier.

A de novo variant is found in a child but in neither parent. It arises in one parent's egg or sperm or just after fertilization, so the child can carry it in every cell with no family history, and can pass it on. If it arose in only some of a parent's egg or sperm cells, an unaffected parent can pass the condition to a child.

Genes on one chromosome tend to enter a gamete in the same combination they came in from a parent, and only crossing over between the paired chromosomes splits them. The closer two genes sit, the less room there is for an exchange between them, so how tightly two genes stay together tells you roughly how close they are.

In a cross that follows two linked genes, offspring carrying the parents' original allele combinations outnumber those carrying new combinations made by crossing over. Genes that assort freely give the two kinds in equal numbers, so a surplus of parental types is the evidence of linkage, and the smaller recombinant classes are the measure of distance.

Divide the recombinant offspring by all offspring and multiply by 100. The figure tops out near 50 percent because that is what genes on different chromosomes give by free assortment, and over long distances double crossovers go uncounted, so the raw value understates the gap. Two genes far apart on one chromosome can therefore look unlinked.

One centimorgan means a 1 percent chance that two markers are separated by recombination in a single meiosis, so it measures how often exchange happens rather than a length. In the human genome it averages about a million base pairs, but crossovers start at preferred hotspots, so the conversion changes from region to region.

Following three linked genes at once gives eight offspring classes, and the two rarest are the double crossovers. Comparing them with the parental classes is how the gene order, and so the middle gene, is worked out. Counting those doubles back in corrects the outer distance, which two-gene crosses understate.

A physical map places landmarks at their actual positions along the DNA, while a genetic map orders them by how often they are inherited together. Because recombination begins preferentially at hotspots instead of evenly along a chromosome, equal genetic distances can cover very different stretches of DNA.

A marker is a DNA landmark at a known spot that differs between people, so you can follow it through a family without any visible trait. Microsatellites vary in repeat count, SNPs at a single base, and RFLPs in whether an enzyme cuts. A marker inherited with a disease points to a gene nearby; it is usually not the gene itself.

A substitution swaps one base for another. A transition keeps the base's chemical class (purine for purine, pyrimidine for pyrimidine); a transversion switches class. There are twice as many possible transversions, yet transitions are the more common change. What a substitution does to a protein is a separate question.

Because several codons can encode the same amino acid, a substitution may leave the protein unchanged (synonymous, or silent), swap one amino acid (missense), or turn a codon into a stop signal (nonsense) and cut the protein short. Missense effects range from harmless to disabling, and a synonymous change is not always without effect.

A cell reads a gene three bases at a time, so adding or removing bases in a count that three does not divide evenly regroups every codon after the change. The downstream amino acids are wrong, and a premature stop codon often appears, which is why such proteins are usually nonfunctional however small the original change.

When an insertion or deletion is a multiple of three bases, the codons after it stay in register, so the protein gains or loses amino acids only at that spot. That makes the effect harder to predict than a frameshift's, not necessarily milder: a deletion of just three bases can still destroy a protein's function.

Some genes contain a short sequence repeated many times in a row, and a stretch with too many repeats is unstable when copied. The count can grow as the gene passes from parent to child, so symptoms may appear earlier or more severely in later generations, a pattern called anticipation, seen in several nervous system disorders.

A germline variant is present in an egg or sperm, so a child who inherits it carries it in virtually every cell and can pass it on. A somatic variant arises later in a body cell and reaches only that cell's descendants; it can affect the person and sometimes leads to disease, but it is not passed to children.

Mutations arise from copying errors during cell division as well as from mutagens such as ultraviolet light, x-rays and many chemicals, so they occur even without any exposure. Most carcinogens cause cancer by mutating DNA, which is why the two lists overlap, but a carcinogen is defined by the disease it causes and a mutagen by the DNA change.

Cells match the repair route to the damage. Mismatch repair fixes wrong pairings left by replication; base excision repair removes single damaged bases; nucleotide excision repair cuts out bulky, helix-distorting lesions such as UV damage. Double-strand breaks go to homologous recombination, which uses the sister chromatid and is accurate, or to end joining, which can lose bases.

Most variants have no effect on health or development, and repair enzymes fix many before they matter. A small share cause disease, and a very small share help, producing a protein that suits a changed environment. Only variants in egg or sperm cells reach the next generation and can feed natural selection.

An allele frequency is the allele's count divided by the total number of copies of that gene in the sample, which is twice the number of people, since each person carries two copies. A homozygote contributes two copies and a heterozygote one. Genotype frequencies describe people instead, and predicting them from allele frequencies alone assumes Hardy-Weinberg conditions.

With allele frequencies p and q, the expected genotype shares are p squared, 2pq and q squared. The expectation holds only with no new mutation, no selection, random mating, no migration and a population large enough that drift does not act. Real populations rarely meet all five, so a gap between observed and expected genotypes is a clue, often to a genotyping error or mixed groups.

For a recessive condition, the share of affected births estimates q squared, so taking its square root gives the allele frequency q, and the carrier frequency is 2pq. Because carriers far outnumber affected people when q is small, a rare condition can still have a common carrier state. The same equation runs backward from carrier counts to a predicted incidence.

Genetic drift is a shift in allele frequencies caused by chance alone, by which copies happen to reach the next generation. It does not depend on whether an allele helps or harms, and it is strongest in small, isolated populations, where it can push a rare allele to high frequency or remove an allele entirely and reduce variation.

Both are sampling accidents: the next generations descend from a small number of individuals taken from a larger population, so they inherit only the variation those few happened to carry. Diversity drops, and a variant that is rare elsewhere can become common in a population descended from few founders.

Gene flow is the movement of alleles between populations through migration and interbreeding. It pulls populations toward the same allele frequencies, working against both drift and local differences, and it can restore diversity a small population has lost. It violates the no-migration assumption of Hardy-Weinberg.

Selection is named for which part of a trait's range wins. Directional selection favors one extreme and shifts the average; stabilizing selection favors the middle and narrows the spread; disruptive selection favors both extremes over the middle and can split one population into distinct forms. Meta-analyses suggest disruptive selection may be at least as common in nature as stabilizing selection.

When carriers of one copy outsurvive both homozygotes, selection keeps a harmful allele in the population. Sickle cell trait cuts the risk of severe falciparum malaria by roughly 90 percent, while two copies cause sickle cell disease, so the allele is most common where that malaria was historically endemic.

Run from 1990 to 2003 by an international consortium, the project aimed to sequence the whole human genome plus several model organisms, including E. coli, yeast, the fruit fly, a nematode and the mouse. Detailed genetic and physical maps were goals along the way. The 2003 sequence covered about 92 percent of the genome; very repetitive stretches stayed unresolved until 2022.

A reference genome is an agreed standard you compare new sequences against, not the DNA of any one person; the original was assembled from several anonymous donors. Because a reference built from a handful of people misses variation found in others, it keeps being revised, and pangenomes now gather many genomes to serve as a broader reference.

When the draft sequence appeared, the human gene count was first put near 35,000, and it was later revised to about 20,000. Protein-coding sequence fills less than 2 percent of the genome; most of the rest is transcribed into non-coding RNA or made of regulatory elements, a part of the genome long dismissed as irrelevant.

Comparing genomes across species finds sequence that has been conserved over millions of years, which flags genes essential to life and signals that control gene activity. About 60 percent of fruit fly genes are conserved in humans, and differences between related genomes help trace how species changed over time.

Model organisms stand in for humans in experiments because they share much of our genetic machinery. Two-thirds of human genes known to be involved in cancer have fruit fly counterparts, and knockout mice, each with one gene switched off, show what a similar human gene may do. A model has limits: about 15 percent of mouse knockouts cannot develop into adults.

A sequence does not tell you what each stretch does. Annotation adds function from evidence: projects such as ENCODE map biochemical activity to catalog candidate regulatory elements, comparison with other species flags conserved sequence, and knocking a gene out in a model organism tests what it is needed for.

At meetings in Bermuda, Human Genome Project scientists agreed to release human sequence publicly within 24 hours of generating it. That choice let any lab use the data at once, and the rapid-release norm became one of the project's most important legacies for biomedical research.

Each cycle runs three steps at different temperatures. Heat denatures the double helix into single strands, cooling lets the two primers anneal to their matching sites, and the polymerase then extends each primer by copying the template. Because each step needs its own temperature, a thermocycler changes the temperature every few minutes.

Every strand made in one cycle serves as a template in the next, which is why 30 to 40 cycles can turn a few starting molecules into more than a billion copies. The same arithmetic works on a single stray molecule, so a trace of contaminating DNA is amplified just as efficiently as the real target.

The two primers set the ends of the product, because copying starts only where each binds. Length matters: longer primers are more unique, but past a point they tolerate mismatches and fold or pair with each other. Primers of about 18 to 24 bases are usually specific when the annealing temperature is matched to their melting temperature, which depends largely on GC content.

The first PCR used a polymerase destroyed by heat, so fresh enzyme had to be added after every denaturing step. Switching to a heat-stable polymerase from the bacterium Thermus aquaticus, which survives the repeated heating, is what made routine, machine-run cycling possible.

The annealing temperature decides how perfect a match a primer needs before it sticks. Set it too low and primers bind loosely matched sites or form structures, giving extra products and lower efficiency. Set it well above the primers' melting temperature and they bind too poorly to give any visible product.

Because PCR copies whatever matching DNA is present, a few contaminating molecules from a reagent, a surface or an earlier reaction can produce a false positive, and the risk grows as the true target gets scarcer. A no-template control, a reaction run without any sample, should stay blank; product there means contamination.

Quantitative PCR adds fluorescent dyes or probes so you can watch product build up cycle by cycle; the cycle at which the signal rises above background reflects how much template the sample started with. To detect an RNA target, such as an RNA virus, reverse transcription first copies the RNA into DNA that PCR can amplify.

Every DNA fragment carries negative charge along its sugar-phosphate backbone, so an electric current can move all of them through a gel. The gel acts as a sieve, letting smaller fragments move faster than larger ones, so after a run the fragments sit in bands ordered by size rather than by charge.

Bands nearest the wells hold the largest fragments; the furthest bands hold the smallest. To turn distance into size, you run a size marker of fragments of known lengths beside the samples and compare each band with it. Migration also depends on gel concentration, voltage and the DNA's shape, which is why the marker goes into every run.

Both gels separate by sieving, and the gel's makeup sets which sizes it can tell apart. A standard agarose gel resolves typical PCR products and digests but cannot separate fragments differing by only a few bases; higher agarose concentrations or polyacrylamide-based methods are used for finer differences. Proteins for a western blot are separated on polyacrylamide.

A restriction enzyme cuts at every copy of its recognition sequence, so digesting a defined DNA molecule always yields the same set of fragments and the same band pattern on a gel. Comparing bands with the sizes you expect checks what the molecule contains, and a single base change that creates or destroys a site changes the pattern.

All three separate molecules on a gel, transfer them to a membrane and then detect one target. A Southern blot finds a DNA sequence with a labeled DNA probe, useful for spotting a deleted or rearranged gene; a Northern blot finds RNA, showing how strongly a gene is expressed; a western blot finds a protein using antibodies and reports its size.

Two single strands with complementary sequences pair into a double strand, and the pairing depends on matching bases. A probe is a labeled single strand designed to pair with one target, so it marks that one sequence among many others, for example on a blot membrane. The label, radioactive, fluorescent or chemical, is what makes the bound probe visible.

Instead of reading a whole genome, a DNA profile looks at a set of microsatellites, short repeats whose number varies between people. Comparing the repeat counts at those sites between two samples shows whether they could come from the same person. A match narrows down who left a sample; it does not by itself show who committed a crime.

Bacteria and archaea make restriction enzymes to cut up invading viral DNA, and they protect their own DNA by methylating the same sites. Many of the enzymes used in labs recognize short palindromic sequences, reading the same on both strands, and cut at a fixed position in or near that site every time.

Some enzymes cut both strands at the same point and leave flush, blunt ends; others cut the strands a few bases apart and leave short single-stranded overhangs. Overhangs from the same enzyme are complementary, so any two pieces cut with it can pair up and be joined, which makes such sticky-end cuts the workhorse of cloning.

Base pairing between overhangs holds two fragments together only loosely; the backbone still has breaks. DNA ligase seals those breaks, turning two DNA pieces from different sources into one continuous molecule. Cutting with one restriction enzyme and then joining with ligase is the step that started recombinant DNA work.

A plasmid is a small circle of DNA that copies itself separately from the bacterial chromosome. A cloning plasmid carries an origin of replication, which also sets how many copies each cell holds; a multiple cloning site where the insert goes; and a selection marker, usually antibiotic resistance, so that, on that antibiotic, cells without the plasmid cannot grow.

Transformation gets a plasmid into bacteria, often by making cells competent with calcium chloride and giving them a brief heat shock. Cells that did not take up the plasmid must be weeded out, so the cells are grown on an antibiotic that the plasmid's resistance gene defeats, and only cells carrying the plasmid form colonies.

To make a protein, the gene must sit behind a promoter the host recognizes, and the host must be able to finish the protein. E. coli is fast and cheap but lacks many eukaryotic post-translational modifications such as proper glycosylation, and its products can clump into inclusion bodies. Yeast can add modifications, and so can mammalian cells, though even their sugar patterns differ from human ones.

Eukaryotic genes are interrupted by introns that a spliceosome removes from the RNA, and bacteria have no spliceosome to do that. Copying a cell's messenger RNA back into DNA with reverse transcription yields coding sequence that is already spliced, so a bacterium can make the protein from it, and the collection reflects the genes that cell was expressing.

Once a human gene is inside bacteria or yeast, the host's machinery copies and expresses it along with its own genes, turning the culture into a factory for a protein used to treat disease, such as growth hormone. For many such medicines, the sugar chains the host attaches affect how well the protein works and how long it lasts in the body.

Sanger sequencing copies a template in the presence of a small amount of labeled dideoxynucleotides, which stop the growing strand wherever one is added. Stops happen at random positions, so the reaction yields fragments ending at every base; sorting them by length and reading each fragment's final labeled base spells out the sequence. It still serves well for short stretches.

Second-generation instruments run huge numbers of sequencing reactions side by side, raising output by orders of magnitude and cutting a whole genome from years of work to days or weeks. The trade-off is read length: reads of about 150 to 300 bases are hard to place in repetitive DNA and lose information about which variants sit on the same chromosome copy.

Long-read methods read single molecules, averaging over 10,000 bases per read, so one read can span a repeat expansion, a large structural rearrangement or two variants that need phasing. Nanopore instruments identify each base by how it changes an electrical current as the strand threads through a tiny pore. The cost is a higher per-read error rate than short reads.

Coverage, or depth, is the average number of reads covering each position. Because any single read can contain an error, a base seen in many independent reads can be called with more confidence than one seen in a few. Assembling a genome also needs enough overlapping reads at every position, and repetitive or duplicated genomes need more data.

De novo assembly rebuilds a genome by stitching overlapping reads together with no reference, which is the only option when no reference exists. Resequencing instead maps each read onto an existing reference to find where an individual differs. Mapping inherits the reference's gaps and errors, and short reads from repeats are hard to place either way.

A variant caller weighs the reads at each position against the chance of sequencing error. An inherited variant is expected in about half the reads (one copy) or all of them (two copies), so the caller asks which of the three genotypes fits the reads best. Repeats and homopolymer runs cause false calls, so callers filter those regions.

BLAST compares a query sequence with a database to find likely homologs, sequences that share evolutionary history. It scores short word matches, extends good ones into alignments, and reports an E-value: the number of hits that well matched you would expect by chance alone in a database that size. The smaller the E-value, the less likely the match is a coincidence.

A gene panel reads a chosen set of genes, the exome reads the roughly 1 percent of the genome made of exons, and whole genome sequencing reads all of it. Exome sequencing catches most known disease-causing changes but misses variants outside exons that affect gene activity. Broader tests find more variants, and the meaning of many of them is unknown.

A genome-wide association study compares people who have a condition with similar people who do not, scanning SNP markers across each genome. A variant that turns up significantly more often in the affected group is said to be associated with the condition. Association marks a region worth studying; it does not show that the variant itself causes anything.

Nearby variants travel through generations in blocks, called haplotypes, because recombination between them is rare. So a SNP flagged in a genome-wide study is often just tagging along with the real causal variant somewhere in its block, and researchers often take further steps, such as sequencing the region, to find the change that actually matters.

Testing about a million independent variants at the usual 0.05 cutoff would flag tens of thousands by chance alone. To limit false positives, genome-wide studies typically require p below 5 times 10 to the minus 8, a Bonferroni correction for a million tests. Some researchers argue the multiple-testing framing does not fit GWAS, yet still back a strict threshold because any one variant has a low prior chance of a true association.

With a large enough sample, even a tiny effect clears the genome-wide threshold. Most variants found this way raise disease odds by about 10 percent or less and typically explain under 1 percent of a trait's variation. Small effects can still be useful: some point to genes that make good drug targets.

Heritability estimates how much of the variation in a trait across a population comes from genetic differences, and it holds only for that population and environment. A heritability of 0.7 does not mean a trait is 70 percent caused by genes in any one person. It also says nothing about which genes are involved or how easily the trait can change: hair color is highly heritable yet easy to dye.

A polygenic risk score adds up the estimated effects of many variants into one number, using genomic information alone and leaving out environment and habits. It shows where a person falls relative to others, but its ability to sort individuals into who will and will not develop a disease is currently low in the general population. Scores also work less well for people whose ancestry differs from the studies that built them.

Many drugs are broken down by enzymes, and variants in the genes for those enzymes can make a person process a drug too fast, so it works poorly, or too slowly, so it builds up and causes side effects. Pharmacogenomic testing helps a prescriber choose among drugs and doses, and FDA drug labels can carry genotype-specific dosing information.

Bacteria that survive a virus store short pieces of its DNA in their own genome as CRISPR arrays, a record of past infections. When the same or a related virus attacks again, they make RNA copies of those pieces, which find the matching viral DNA and guide Cas9 or a similar enzyme to cut it. Researchers adapted that defense into a tool for editing DNA.

The guide RNA carries a stretch of about 20 bases matching the target, and it also binds Cas9, steering the enzyme to that spot by base pairing. Retargeting means writing a new guide sequence rather than designing a new protein, one reason CRISPR is simpler, faster and cheaper than older editing methods.

Cas9 first scans DNA for a short protospacer adjacent motif (PAM) next to the target, and only then unwinds the DNA so the guide can pair. The widely used Cas9 from Streptococcus pyogenes needs NGG, where N is any base, so a site lacking a nearby PAM cannot be cut even with a perfect guide. Engineered Cas9 variants relax the requirement.

Cas9 only makes the cut; the cell's repair decides the edit. Non-homologous end joining, the usual route, rejoins the ends with random small insertions or deletions that often shift the reading frame and knock the gene out. Homology-directed repair copies a supplied template to write in a precise change, but it is less efficient and largely limited to dividing cells.

Cas9 can tolerate up to about three mismatches between its guide and the DNA, so it may cut similar sequences elsewhere in the genome. Off-target edits, along with mosaicism, where only some cells carry the edit, are the main safety concerns. Prediction software, high-fidelity Cas9 variants and short-lived delivery all aim to reduce them.

Base editors chemically convert one base into another, C-G to T-A or A-T to G-C, without cutting both DNA strands. Prime editing goes further, installing any of the twelve possible single-base changes plus small insertions or deletions, again with no double-strand break and no donor template, which avoids the random indels that cutting invites.

Zinc-finger nucleases and TALENs find their targets through protein domains that bind DNA in a sequence-specific way, so each new target needs a newly designed protein. Designing zinc fingers was complicated and often inefficient; TALENs were more flexible because each repeat reads one base. TALENs were used in an early somatic editing treatment for leukemia.

Most editing targets somatic cells, so the change stays in the treated person. Editing an embryo, egg or sperm would pass the change to future generations who cannot consent, which is why most ethical debate centers on the germline. NIH does not fund gene editing in human embryos, and bioethicists and researchers generally hold that editing for reproduction should not be attempted at this time.

Gene cloning inserts a piece of DNA into a vector, such as a plasmid, and lets the vector multiply so the gene is copied many times for study; it is the most common kind of cloning NHGRI researchers do. Reproductive cloning makes a whole animal with the same nuclear DNA as a donor. The two share a word and almost none of their methods.

The nucleus of a mature body cell, such as a skin cell, is moved into an egg whose own nucleus has been removed, either by injection or by fusing the two cells with an electric current. The egg develops into an early embryo that a surrogate carries to term. Dolly the sheep, the first mammal cloned from an adult cell, took 276 attempts.

Both use the same nuclear-transfer techniques; the difference is the goal. Reproductive cloning implants the embryo to produce an animal, while therapeutic cloning stops at the early embryo to harvest embryonic stem cells for growing replacement tissue. Even a reproductive clone is not a perfect copy, since environment and random processes shape it too.

Potency narrows with development. A totipotent cell, such as the fertilized egg, can form both the embryo and the extra-embryonic tissues that support it. Pluripotent cells, including embryonic stem cells, can become any cell of the adult body but not those outside tissues, and multipotent adult stem cells make only the cell types of their own tissue.

Since 2006, mature cells such as skin cells can be reprogrammed into an embryonic-like pluripotent state by forcing the expression of genes that act as master regulators of pluripotency over several weeks. Because no embryo is needed, induced pluripotent stem cells avoid that ethical objection, and cells made from a patient may avoid immune rejection.

DNA dropped into a cell usually does nothing, so gene therapy relies on a vector, often a disabled virus. Retroviruses insert their cargo into a chromosome, while adenoviruses deliver DNA that stays outside the chromosomes. The vector can be given directly to the body, or cells can be removed, treated in the lab and returned. Early trials ran into toxicity, inflammation and cancer, which drove the search for safer methods.

Synthetic biology tries to build organisms the way engineers build machines, from standardized, shared parts such as promoters, ribosome binding sites, coding sequences and terminators, often to make a medicine or fuel or to sense something. The catch is that a part's behavior can change with its surroundings, so designs that work on paper do not always work in a cell.

Most GMO corn is built to resist insect pests or tolerate herbicides: Bt corn makes proteins toxic to certain insects, while herbicide-tolerant canola and alfalfa let farmers spray weeds. Other crops resist viruses, and rice has been engineered to make beta-carotene. In the US, the FDA regulates GMO foods, the EPA regulates the pest-protecting substances such plants make, and USDA APHIS guards against harm to other plants.

A transgenic organism carries DNA introduced from another species, while genome editing typically makes smaller changes to the organism's own DNA. US labeling turns on a related test: the USDA's National Bioengineered Food Disclosure Standard covers foods with detectable genetic material modified in ways that could not come from conventional breeding or occur in nature.

Instead of waiting for a plant to show a trait, breeders test young plants for a DNA marker tightly linked to the gene they want and keep only the carriers. The marker guides which plants to select from existing variation; nothing foreign is inserted. Markers flanking the gene make the prediction more reliable, and the approach has shortened the time to bring new varieties to market.

The FBI's CODIS system compares a set of microsatellite markers from crime-scene DNA with profiles of convicted offenders, arrestees and other scenes, and a database hit is followed by a confirmatory sample from the suspect. A match statistic says how rare a profile is, not how likely that person is to be guilty, which also depends on the rest of the evidence.

Direct-to-consumer tests are sold straight to customers and report traits, ancestry, drug-response predictions and some disease risks without a health care provider. They often check only a few of the variants behind a condition and ignore lifestyle and environment, so a reassuring result can mislead, and results usually need confirmation by clinical testing before they are used for a diagnosis.

Apart from identical twins, every person's DNA sequence is unique, so a genomic sample can never be truly anonymized. A 2013 study re-identified research participants by pairing de-identified genomic data with genealogy databases and public records. Because relatives share DNA, one person's results also say something about family members who never agreed to be tested.

The US Genetic Information Nondiscrimination Act (GINA) bars health insurers from using genetic information to decide eligibility or coverage, and bars employers from using it in hiring, promotion and other job decisions. The gaps are real: GINA does not apply to employers with fewer than 15 employees, and it does not cover life, disability or long-term care insurance.

Exome and genome sequencing can reveal variants linked to conditions other than the one being tested, called secondary or incidental findings. A professional society recommends reporting a set list of genes where early action can help, while variants of uncertain significance are not reported, and people can opt out. Consent therefore has to cover what may be found and what will be returned.

NHGRI set up its Ethical, Legal, and Social Implications research program in 1990, at the start of the Human Genome Project, to study issues such as misuse of genetic information in insurance and employment as the science advanced. Congress later required NHGRI to devote at least 5 percent of its research budget to this work.