Spontaneous mutation

Spontaneous mutations occur with non-zero probability even given a healthy, uncontaminated cell. Naturally occurring oxidative DNA damage is estimated to occur 10,000 times per cell per day in humans and 100,000 times per cell per day in rats.^[31] Spontaneous mutations can be characterized by the specific change:^[32]

• Tautomerism – A base is changed by the repositioning of a hydrogen atom, altering the hydrogen bonding pattern of that base, resulting in incorrect base pairing during replication.^[33] Theoretical results suggest that proton tunneling is an important factor in the spontaneous creation of GC tautomers.^[34]
• Depurination – Loss of a purine base (A or G) to form an apurinic site (AP site).
• Deamination – Hydrolysis changes a normal base to an atypical base containing a keto group in place of the original amine group. Examples include C → U and A → HX (hypoxanthine), which can be corrected by DNA repair mechanisms; and 5MeC (5-methylcytosine) → T, which is less likely to be detected as a mutation because thymine is a normal DNA base.
• Slipped strand mispairing – Denaturation of the new strand from the template during replication, followed by renaturation in a different spot ("slipping"). This can lead to insertions or deletions.

Error-prone replication bypass

There is increasing evidence that the majority of spontaneously arising mutations are due to error-prone replication (translesion synthesis) past DNA damage in the template strand. In mice, the majority of mutations are caused by translesion synthesis.^[35] Likewise, in yeast, Kunz et al.^[36] found that more than 60% of the spontaneous single base pair substitutions and deletions were caused by translesion synthesis.

Errors introduced during DNA repair

Although naturally occurring double-strand breaks occur at a relatively low frequency in DNA, their repair often causes mutation. Non-homologous end joining (NHEJ) is a major pathway for repairing double-strand breaks. NHEJ involves removal of a few nucleotides to allow somewhat inaccurate alignment of the two ends for rejoining followed by addition of nucleotides to fill in gaps. As a consequence, NHEJ often introduces mutations.^[37]

A covalent adduct between the metabolite of benzo[a]pyrene, the major mutagen in tobacco smoke, and DNA^[38]

Induced mutation

Induced mutations are alterations in the gene after it has come in contact with mutagens and environmental causes.

Induced mutations on the molecular level can be caused by:

• Chemicals
  • Hydroxylamine
  • Base analogs (e.g., Bromodeoxyuridine (BrdU))
  • Alkylating agents (e.g., N-ethyl-N-nitrosourea (ENU). These agents can mutate both replicating and non-replicating DNA. In contrast, a base analog can mutate the DNA only when the analog is incorporated in replicating the DNA. Each of these classes of chemical mutagens has certain effects that then lead to transitions, transversions, or deletions.
  • Agents that form DNA adducts (e.g., ochratoxin A)^[39]
  • DNA intercalating agents (e.g., ethidium bromide)
  • DNA crosslinkers
  • Oxidative damage
  • Nitrous acid converts amine groups on A and C to diazo groups, altering their hydrogen bonding patterns, which leads to incorrect base pairing during replication.
• Radiation
  • Ultraviolet light (UV) (including non-ionizing radiation). Two nucleotide bases in DNA—cytosine and thymine—are most vulnerable to radiation that can change their properties. UV light can induce adjacent pyrimidine bases in a DNA strand to become covalently joined as a pyrimidine dimer. UV radiation, in particular longer-wave UVA, can also cause oxidative damage to DNA.^[40]
  • Ionizing radiation. Exposure to ionizing radiation, such as gamma radiation, can result in mutation, possibly resulting in cancer or death.

Whereas in former times mutations were assumed to occur by chance, or induced by mutagens, molecular mechanisms of mutation have been discovered in bacteria and across the tree of life. As S. Rosenberg states, "These mechanisms reveal a picture of highly regulated mutagenesis, up-regulated temporally by stress responses and activated when cells/organisms are maladapted to their environments—when stressed—potentially accelerating adaptation."^[41] Since they are self-induced mutagenic mechanisms that increase the adaptation rate of organisms, they have some times been named as adaptive mutagenesis mechanisms, and include the SOS response in bacteria,^[42] ectopic intrachromosomal recombination^[43] and other chromosomal events such as duplications.^[41]

By effect on structure

Five types of chromosomal mutations

Types of small-scale mutations

The sequence of a gene can be altered in a number of ways.^[44] Gene mutations have varying effects on health depending on where they occur and whether they alter the function of essential proteins. Mutations in the structure of genes can be classified into several types.

By impact on protein sequence

The structure of a eukaryotic protein-coding gene. A mutation in the protein coding region (red) can result in a change in the amino acid sequence. Mutations in other areas of the gene can have diverse effects. Changes within regulatory sequences (yellow and blue) can effect transcriptional and translational regulation of gene expression.

Point mutations classified by impact on protein

Selection of disease-causing mutations, in a standard table of the genetic code of amino acids^[47]

The effect of a mutation on protein sequence depends in part on where in the genome it occurs, especially whether it is in a coding or non-coding region. Mutations in the non-coding regulatory sequences of a gene, such as promoters, enhancers, and silencers, can alter levels of gene expression, but are less likely to alter the protein sequence. Mutations within introns and in regions with no known biological function (e.g. pseudogenes, retrotransposons) are generally neutral, having no effect on phenotype – though intron mutations could alter the protein product if they affect mRNA splicing.

Mutations that occur in coding regions of the genome are more likely to alter the protein product, and can be categorized by their effect on amino acid sequence:

• A frameshift mutation is caused by insertion or deletion of a number of nucleotides that is not evenly divisible by three from a DNA sequence. Due to the triplet nature of gene expression by codons, the insertion or deletion can disrupt the reading frame, or the grouping of the codons, resulting in a completely different translation from the original.^[48] The earlier in the sequence the deletion or insertion occurs, the more altered the protein produced is. (For example, the code CCU GAC UAC CUA codes for the amino acids proline, aspartic acid, tyrosine, and leucine. If the U in CCU was deleted, the resulting sequence would be CCG ACU ACC UAx, which would instead code for proline, threonine, threonine, and part of another amino acid or perhaps a stop codon (where the x stands for the following nucleotide).) By contrast, any insertion or deletion that is evenly divisible by three is termed an in-frame mutation.
• A point substitution mutation results in a change in a single nucleotide and can be either synonymous or nonsynonymous.
  • A synonymous substitution replaces a codon with another codon that codes for the same amino acid, so that the produced amino acid sequence is not modified. Synonymous mutations occur due to the degenerate nature of the genetic code. If this mutation does not result in any phenotypic effects, then it is called silent, but not all synonymous substitutions are silent. (There can also be silent mutations in nucleotides outside of the coding regions, such as the introns, because the exact nucleotide sequence is not as crucial as it is in the coding regions, but these are not considered synonymous substitutions.)
  • A nonsynonymous substitution replaces a codon with another codon that codes for a different amino acid, so that the produced amino acid sequence is modified. Nonsynonymous substitutions can be classified as nonsense or missense mutations:
    • A missense mutation changes a nucleotide to cause substitution of a different amino acid. This in turn can render the resulting protein nonfunctional. Such mutations are responsible for diseases such as Epidermolysis bullosa, sickle-cell disease, and SOD1-mediated ALS.^[49] On the other hand, if a missense mutation occurs in an amino acid codon that results in the use of a different, but chemically similar, amino acid, then sometimes little or no change is rendered in the protein. For example, a change from AAA to AGA will encode arginine, a chemically similar molecule to the intended lysine. In this latter case the mutation will have little or no effect on phenotype and therefore be neutral.
    • A nonsense mutation is a point mutation in a sequence of DNA that results in a premature stop codon, or a nonsense codon in the transcribed mRNA, and possibly a truncated, and often nonfunctional protein product. This sort of mutation has been linked to different diseases, such as congenital adrenal hyperplasia. (See Stop codon.)

By effect on function

A mutation becomes an effect on function mutation when the exactitude of functions between a mutated protein and its direct interactor undergoes change. The interactors can be other proteins, molecules, nucleic acids, etc. There are many mutations that fall under the category of by effect on function, but depending on the specificity of the change the mutations listed below will occur.^[50]

• Loss-of-function mutations, also called inactivating mutations, result in the gene product having less or no function (being partially or wholly inactivated). When the allele has a complete loss of function (null allele), it is often called an amorph or amorphic mutation in Muller's morphs schema. Phenotypes associated with such mutations are most often recessive. Exceptions are when the organism is haploid, or when the reduced dosage of a normal gene product is not enough for a normal phenotype (this is called haploinsufficiency). A disease that is caused by a loss-of-function mutation is Gitelman syndrome and cystic fibrosis.^[51]
• Gain-of-function mutations also called activating mutations, change the gene product such that its effect gets stronger (enhanced activation) or even is superseded by a different and abnormal function. When the new allele is created, a heterozygote containing the newly created allele as well as the original will express the new allele; genetically this defines the mutations as dominant phenotypes. Several of Muller's morphs correspond to the gain of function, including hypermorph (increased gene expression) and neomorph (novel function).
• Dominant negative mutations (also called anti-morphic mutations) have an altered gene product that acts antagonistically to the wild-type allele. These mutations usually result in an altered molecular function (often inactive) and are characterized by a dominant or semi-dominant phenotype. In humans, dominant negative mutations have been implicated in cancer (e.g., mutations in genes p53, ATM, CEBPA, and PPARgamma). Marfan syndrome is caused by mutations in the FBN1 gene, located on chromosome 15, which encodes fibrillin-1, a glycoprotein component of the extracellular matrix. Marfan syndrome is also an example of dominant negative mutation and haploinsufficiency.
• Lethal mutations result in rapid organismal death when occurring during development and cause significant reductions of life expectancy for developed organisms. An example of a disease that is caused by a dominant lethal mutation is Huntington's disease.
• Null mutations, also known as Amorphic mutations, are a form of loss-of-function mutations that completely prohibit the gene's function. The mutation leads to a complete loss of operation at the phenotypic level, also causing no gene product to be formed. Atopic eczema and dermatitis syndrome are common diseases caused by a null mutation of the gene that activates filaggrin.
• Suppressor mutations are a type of mutation that causes the double mutation to appear normally. In suppressor mutations the phenotypic activity of a different mutation is completely suppressed, thus causing the double mutation to look normal. There are two types of suppressor mutations, there are intragenic and extragenic suppressor mutations. Intragenic mutations occur in the gene where the first mutation occurs, while extragenic mutations occur in the gene that interacts with the product of the first mutation. A common disease that results from this type of mutation is Alzheimer's disease.^[52]
• Neomorphic mutations are a part of the gain-of-function mutations and are characterized by the control of new protein product synthesis. The newly synthesized gene normally contains a novel gene expression or molecular function. The result of the neomorphic mutation is the gene where the mutation occurs has a complete change in function.^[53]
• A back mutation or reversion is a point mutation that restores the original sequence and hence the original phenotype.^[54]

By effect on fitness (harmful, beneficial, neutral mutations)

In genetics, it is sometimes useful to classify mutations as either harmful or beneficial (or neutral):

• A harmful, or deleterious, mutation decreases the fitness of the organism. Many, but not all mutations in essential genes are harmful (if a mutation does not change the amino acid sequence in an essential protein, it is harmless in most cases).
• A beneficial, or advantageous mutation increases the fitness of the organism. Examples are mutations that lead to antibiotic resistance in bacteria (which are beneficial for bacteria but usually not for humans).
• A neutral mutation has no harmful or beneficial effect on the organism. Such mutations occur at a steady rate, forming the basis for the molecular clock. In the neutral theory of molecular evolution, neutral mutations provide genetic drift as the basis for most variation at the molecular level. In animals or plants, most mutations are neutral, given that the vast majority of their genomes is either non-coding or consists of repetitive sequences that have no obvious function ("junk DNA").^[55]

Large-scale quantitative mutagenesis screens, in which thousands of millions of mutations are tested, invariably find that a larger fraction of mutations has harmful effects but always returns a number of beneficial mutations as well. For instance, in a screen of all gene deletions in E. coli, 80% of mutations were negative, but 20% were positive, even though many had a very small effect on growth (depending on condition).^[56] Gene deletions involve removal of whole genes, so that point mutations almost always have a much smaller effect. In a similar screen in Streptococcus pneumoniae, but this time with transposon insertions, 76% of insertion mutants were classified as neutral, 16% had a significantly reduced fitness, but 6% were advantageous.^[57]

This classification is obviously relative and somewhat artificial: a harmful mutation can quickly turn into a beneficial mutations when conditions change. Also, there is a gradient from harmful/beneficial to neutral, as many mutations may have small and mostly neglectable effects but under certain conditions will become relevant. Also, many traits are determined by hundreds of genes (or loci), so that each locus has only a minor effect. For instance, human height is determined by hundreds of genetic variants ("mutations") but each of them has a very minor effect on height,^[58] apart from the impact of nutrition. Height (or size) itself may be more or less beneficial as the huge range of sizes in animal or plant groups shows.

Distribution of fitness effects (DFE)

Attempts have been made to infer the distribution of fitness effects (DFE) using mutagenesis experiments and theoretical models applied to molecular sequence data. DFE, as used to determine the relative abundance of different types of mutations (i.e., strongly deleterious, nearly neutral or advantageous), is relevant to many evolutionary questions, such as the maintenance of genetic variation,^[59] the rate of genomic decay,^[60] the maintenance of outcrossing sexual reproduction as opposed to inbreeding^[61] and the evolution of sex and genetic recombination.^[62] DFE can also be tracked by tracking the skewness of the distribution of mutations with putatively severe effects as compared to the distribution of mutations with putatively mild or absent effect.^[63] In summary, the DFE plays an important role in predicting evolutionary dynamics.^[64][65] A variety of approaches have been used to study the DFE, including theoretical, experimental and analytical methods.

• Mutagenesis experiment: The direct method to investigate the DFE is to induce mutations and then measure the mutational fitness effects, which has already been done in viruses, bacteria, yeast, and Drosophila. For example, most studies of the DFE in viruses used site-directed mutagenesis to create point mutations and measure relative fitness of each mutant.^{[66][67][68][69]} In Escherichia coli, one study used transposon mutagenesis to directly measure the fitness of a random insertion of a derivative of Tn10.^[70] In yeast, a combined mutagenesis and deep sequencing approach has been developed to generate high-quality systematic mutant libraries and measure fitness in high throughput.^[71] However, given that many mutations have effects too small to be detected^[72] and that mutagenesis experiments can detect only mutations of moderately large effect; DNA sequence analysis can provide valuable information about these mutations.

The distribution of fitness effects (DFE) of mutations in vesicular stomatitis virus. In this experiment, random mutations were introduced into the virus by site-directed mutagenesis, and the fitness of each mutant was compared with the ancestral type. A fitness of zero, less than one, one, more than one, respectively, indicates that mutations are lethal, deleterious, neutral, and advantageous.^[66]

• 

  This figure shows a simplified version of loss-of-function, switch-of-function, gain-of-function, and conservation-of-function mutations.
  Molecular sequence analysis: With rapid development of DNA sequencing technology, an enormous amount of DNA sequence data is available and even more is forthcoming in the future. Various methods have been developed to infer the DFE from DNA sequence data.^{[73][74][75][76]} By examining DNA sequence differences within and between species, we are able to infer various characteristics of the DFE for neutral, deleterious and advantageous mutations.^[24] To be specific, the DNA sequence analysis approach allows us to estimate the effects of mutations with very small effects, which are hardly detectable through mutagenesis experiments.

One of the earliest theoretical studies of the distribution of fitness effects was done by Motoo Kimura, an influential theoretical population geneticist. His neutral theory of molecular evolution proposes that most novel mutations will be highly deleterious, with a small fraction being neutral.^[25][77] A later proposal by Hiroshi Akashi proposed a bimodal model for the DFE, with modes centered around highly deleterious and neutral mutations.^[78] Both theories agree that the vast majority of novel mutations are neutral or deleterious and that advantageous mutations are rare, which has been supported by experimental results. One example is a study done on the DFE of random mutations in vesicular stomatitis virus.^[66] Out of all mutations, 39.6% were lethal, 31.2% were non-lethal deleterious, and 27.1% were neutral. Another example comes from a high throughput mutagenesis experiment with yeast.^[71] In this experiment it was shown that the overall DFE is bimodal, with a cluster of neutral mutations, and a broad distribution of deleterious mutations.

Though relatively few mutations are advantageous, those that are play an important role in evolutionary changes.^[79] Like neutral mutations, weakly selected advantageous mutations can be lost due to random genetic drift, but strongly selected advantageous mutations are more likely to be fixed. Knowing the DFE of advantageous mutations may lead to increased ability to predict the evolutionary dynamics. Theoretical work on the DFE for advantageous mutations has been done by John H. Gillespie^[80] and H. Allen Orr.^[81] They proposed that the distribution for advantageous mutations should be exponential under a wide range of conditions, which, in general, has been supported by experimental studies, at least for strongly selected advantageous mutations.^[82][83][84]

In general, it is accepted that the majority of mutations are neutral or deleterious, with advantageous mutations being rare; however, the proportion of types of mutations varies between species. This indicates two important points: first, the proportion of effectively neutral mutations is likely to vary between species, resulting from dependence on effective population size; second, the average effect of deleterious mutations varies dramatically between species.^[24] In addition, the DFE also differs between coding regions and noncoding regions, with the DFE of noncoding DNA containing more weakly selected mutations.^[24]

By inheritance

A mutation has caused this moss rose plant to produce flowers of different colors. This is a somatic mutation that may also be passed on in the germline.

In multicellular organisms with dedicated reproductive cells, mutations can be subdivided into germline mutations, which can be passed on to descendants through their reproductive cells, and somatic mutations (also called acquired mutations),^[85] which involve cells outside the dedicated reproductive group and which are not usually transmitted to descendants.

Diploid organisms (e.g., humans) contain two copies of each gene—a paternal and a maternal allele. Based on the occurrence of mutation on each chromosome, we may classify mutations into three types. A wild type or homozygous non-mutated organism is one in which neither allele is mutated.

• A heterozygous mutation is a mutation of only one allele.
• A homozygous mutation is an identical mutation of both the paternal and maternal alleles.
• Compound heterozygous mutations or a genetic compound consists of two different mutations in the paternal and maternal alleles.^[86]

Special classes

• Conditional mutation is a mutation that has wild-type (or less severe) phenotype under certain "permissive" environmental conditions and a mutant phenotype under certain "restrictive" conditions. For example, a temperature-sensitive mutation can cause cell death at high temperature (restrictive condition), but might have no deleterious consequences at a lower temperature (permissive condition).^[91] These mutations are non-autonomous, as their manifestation depends upon presence of certain conditions, as opposed to other mutations which appear autonomously.^[92] The permissive conditions may be temperature,^[93] certain chemicals,^[94] light^[94] or mutations in other parts of the genome.^[92] In vivo mechanisms like transcriptional switches can create conditional mutations. For instance, association of Steroid Binding Domain can create a transcriptional switch that can change the expression of a gene based on the presence of a steroid ligand.^[95] Conditional mutations have applications in research as they allow control over gene expression. This is especially useful studying diseases in adults by allowing expression after a certain period of growth, thus eliminating the deleterious effect of gene expression seen during stages of development in model organisms.^[94] DNA Recombinase systems like Cre-Lox recombination used in association with promoters that are activated under certain conditions can generate conditional mutations. Dual Recombinase technology can be used to induce multiple conditional mutations to study the diseases which manifest as a result of simultaneous mutations in multiple genes.^[94] Certain inteins have been identified which splice only at certain permissive temperatures, leading to improper protein synthesis and thus, loss-of-function mutations at other temperatures.^[96] Conditional mutations may also be used in genetic studies associated with ageing, as the expression can be changed after a certain time period in the organism's lifespan.^[93]
• Replication timing quantitative trait loci affects DNA replication.

Nomenclature

In order to categorize a mutation as such, the "normal" sequence must be obtained from the DNA of a "normal" or "healthy" organism (as opposed to a "mutant" or "sick" one), it should be identified and reported; ideally, it should be made publicly available for a straightforward nucleotide-by-nucleotide comparison, and agreed upon by the scientific community or by a group of expert geneticists and biologists, who have the responsibility of establishing the standard or so-called "consensus" sequence. This step requires a tremendous scientific effort. Once the consensus sequence is known, the mutations in a genome can be pinpointed, described, and classified. The committee of the Human Genome Variation Society (HGVS) has developed the standard human sequence variant nomenclature,^[97] which should be used by researchers and DNA diagnostic centers to generate unambiguous mutation descriptions. In principle, this nomenclature can also be used to describe mutations in other organisms. The nomenclature specifies the type of mutation and base or amino acid changes.

• Nucleotide substitution (e.g., 76A>T) – The number is the position of the nucleotide from the 5' end; the first letter represents the wild-type nucleotide, and the second letter represents the nucleotide that replaced the wild type. In the given example, the adenine at the 76th position was replaced by a thymine.
  • If it becomes necessary to differentiate between mutations in genomic DNA, mitochondrial DNA, and RNA, a simple convention is used. For example, if the 100th base of a nucleotide sequence mutated from G to C, then it would be written as g.100G>C if the mutation occurred in genomic DNA, m.100G>C if the mutation occurred in mitochondrial DNA, or r.100g>c if the mutation occurred in RNA. Note that, for mutations in RNA, the nucleotide code is written in lower case.
• Amino acid substitution (e.g., D111E) – The first letter is the one letter code of the wild-type amino acid, the number is the position of the amino acid from the N-terminus, and the second letter is the one letter code of the amino acid present in the mutation. Nonsense mutations are represented with an X for the second amino acid (e.g. D111X).
• Amino acid deletion (e.g., ΔF508) – The Greek letter Δ (delta) indicates a deletion. The letter refers to the amino acid present in the wild type and the number is the position from the N terminus of the amino acid were it to be present as in the wild type.

Randomness of mutations

There is a widespread assumption that mutations are (entirely) "random" with respect to their consequences (in terms of probability). This was shown to be wrong as mutation frequency can vary across regions of the genome, with such DNA repair- and mutation-biases being associated with various factors. For instance, biologically important regions were found to be protected from mutations and mutations beneficial to the studied plant were found to be more likely – i.e. mutation is "non-random in a way that benefits the plant".^[105][106] Additionally, previous experiments typically used to demonstrate mutations being random with respect to fitness (such as the Fluctuation Test and Replica plating) have been shown to only support the weaker claim that those mutations are random with respect to external selective constraints, not fitness as a whole.^[107]

Inherited disorders

If a mutation is present in a germ cell, it can give rise to offspring that carries the mutation in all of its cells. This is the case in hereditary diseases. In particular, if there is a mutation in a DNA repair gene within a germ cell, humans carrying such germline mutations may have an increased risk of cancer. A list of 34 such germline mutations is given in the article DNA repair-deficiency disorder. An example of one is albinism, a mutation that occurs in the OCA1 or OCA2 gene. Individuals with this disorder are more prone to many types of cancers, other disorders and have impaired vision.

DNA damage can cause an error when the DNA is replicated, and this error of replication can cause a gene mutation that, in turn, could cause a genetic disorder. DNA damages are repaired by the DNA repair system of the cell. Each cell has a number of pathways through which enzymes recognize and repair damages in DNA. Because DNA can be damaged in many ways, the process of DNA repair is an important way in which the body protects itself from disease. Once DNA damage has given rise to a mutation, the mutation cannot be repaired.

Role in carcinogenesis

On the other hand, a mutation may occur in a somatic cell of an organism. Such mutations will be present in all descendants of this cell within the same organism. The accumulation of certain mutations over generations of somatic cells is part of cause of malignant transformation, from normal cell to cancer cell.^[109]

Cells with heterozygous loss-of-function mutations (one good copy of gene and one mutated copy) may function normally with the unmutated copy until the good copy has been spontaneously somatically mutated. This kind of mutation happens often in living organisms, but it is difficult to measure the rate. Measuring this rate is important in predicting the rate at which people may develop cancer.^[110]

Point mutations may arise from spontaneous mutations that occur during DNA replication. The rate of mutation may be increased by mutagens. Mutagens can be physical, such as radiation from UV rays, X-rays or extreme heat, or chemical (molecules that misplace base pairs or disrupt the helical shape of DNA). Mutagens associated with cancers are often studied to learn about cancer and its prevention.

Prion mutations

Prions are proteins and do not contain genetic material. However, prion replication has been shown to be subject to mutation and natural selection just like other forms of replication.^[117] The human gene PRNP codes for the major prion protein, PrP, and is subject to mutations that can give rise to disease-causing prions.

Compensatory mutations in RNA

As the function of a RNA molecule is dependent on its structure,^[123] the structure of RNA molecules is evolutionarily conserved. Therefore, any mutation that alters the stable structure of RNA molecules must be compensated by other compensatory mutations. In the context of RNA, the sequence of the RNA can be considered as ' genotype' and the structure of the RNA can be considered as its 'phenotype'. Since RNAs have relatively simpler composition than proteins, the structure of RNA molecules can be computationally predicted with high degree of accuracy. Because of this convenience, compensatory mutations have been studied in computational simulations using RNA folding algorithms.^[124][125]

Evolutionary mechanism of compensation

Compensatory mutations can be explained by the genetic phenomenon epistasis whereby the phenotypic effect of one mutation is dependent upon mutation(s) at other loci. While epistasis was originally conceived in the context of interaction between different genes, intragenic epistasis has also been studied recently.^[126] Existence of compensated pathogenic deviations can be explained by 'sign epistasis', in which the effects of a deleterious mutation can be compensated by the presence of an epistatic mutation in another loci. For a given protein, a deleterious mutation (D) and a compensatory mutation (C) can be considered, where C can be in the same protein as D or in a different interacting protein depending on the context. The fitness effect of C itself could be neutral or somewhat deleterious such that it can still exist in the population, and the effect of D is deleterious to the extent that it cannot exist in the population. However, when C and D co-occur together, the combined fitness effect becomes neutral or positive.^[120] Thus, compensatory mutations can bring novelty to proteins by forging new pathways of protein evolution : it allows individuals to travel from one fitness peak to another through the valleys of lower fitness.^[126] 

DePristo et al. 2005 outlined two models to explain the dynamics of compensatory pathogenic deviations (CPD).^[127] In the first hypothesis P is a pathogenic amino acid mutation that and C is a neutral compensatory mutation.^[127] Under these conditions, if the pathogenic mutation arises after a compensatory mutation, then P can become fixed in the population.^[127] The second model of CPDs states that P and C are both deleterious mutations resulting in fitness valleys when mutations occur simultaneously.^[127] Using publicly available, Ferrer-Costa et al. 2007 obtained compensatory mutations and human pathogenic mutation datasets that were characterized to determine what causes CPDs.^[128] Results indicate that the structural constraints and the location in protein structure determine whether compensated mutations will occur.^[128]

Experimental evidence of compensatory mutations