Correlated Characters

Correlated Characters

Characters are often correlated, that is, the phenotypic value of one character in an individual is correlated with the phenotypic value of another character on that individual. (The circumference of your head is about 1/3 of your height.)

These correlations can also be due to environmental effects or genetic effects. The genetic causes of correlation are pleiotropy (that genes affect more than one character) and linkage disequilibrium.

This need not be constant across genes: some genes can cause positive pleiotropy and others negative pleiotropy; the balance determines the genetic correlation of the two characters.

More r’s!! Correlations are denoted by r, with subscripts to indicate genetic, environmental, phenotypic correlations.

The sign of the value of the genetic correlation can positive or negative, and need not be the same as the phenotypic correlation.

Definition, again

A correlation is the covariance divided by the standard deviations of the two things.

Estimation

Estimation of genetic correlations is done by analysis of covariance, where the values of one relative for one trait are correlated with the values of another relative for a different trait.

These estimates generally proceed in much the same way as the analysis of variance components, with the same coefficients for the same class of relatives.

The estimates of correlations, being ratios, are notoriously imprecise.

Correlated response to selection

Selection on one trait will often result in response for another trait. This is genetic correlations. It is caused by the changes in the breeding value of the selected trait being correlated with changes in the breeding value of the other trait.

CR_Y = b_(A)YX R_X =

Selection on one trait can cause an apparent selection differential at another trait, because of both the genetic and environmental correlations. This is a particularly huge problem when studying natural selection in natural conditions.

Indirect selection

Sometimes it is easier to get a response to selection for a trait by selecting on a correlated trait instead. This is the case when the heritability for the secondary trait is smaller or when it is easier or cheaper to measure.

GxE and genetic correlations

A great way to consider GxE is that the expression of a character in different environments can be considered different characters, with some genetic correlations. We’ll get to this more in the lecture on phenotypic plasticity (November 14).

QTL’s
Quantitative Trait Loci = QTL

Any locus which has genetic variation which affects a particular trait is QTL.

These loci are easier to identify if their effects on the phenotype are large; in this case they are also referred to as a major gene.

In practice, what are identified as QTL are often not truly single loci in the standard sense, but rather chromosomal regions. The distinction can only be made in special genetic circumstances.

There are many examples of major genes.

Remember pygmy from problem 6.2?

How to discover major genes:
Get lucky
Observation of segregation patterns
Correlation with markers
Correlation with "candidate loci" variation

These last two are the basis of QTL analysis.

Mapping QTL’s
Need map of the markers
Marker loci:
Need to be highly polymorphic
Abundant and evenly spread
neutral
hopefully co-dominant, although Zeng has shown that there is also a lot of info in recessive markers

e.g. RFLP’s, microsatellites, transposon insertion sites, or (less useful) RAPD’s
The strength of any QTL analysis depends on the extent of linkage disequilibrium
Markers must be in LD with QTL alleles
Preferably, QTL alleles are also in association, that is, that alleles which increase a phenotypes’ value are in linkage disequilibrium with one another.

The process in general:

Therefore, inbred parental strains are often used. Also, up and down selection lines are often compared.

F1 crosses between the parental lines are then used to create F1 and/or backcross lines for comparison.

All individuals used must then be scored for both phenotype and marker genotype --very time-consuming and expensive

Correlations between phenotype and the marker alleles indicate genetic factors linked to the markers

Randomly bred individuals give very, very little information, although techniques for this case are being developed (Ritland 1996 Evolution)

Caveats

The estimates are always an underestimate, since it is only chromosomal regions which are investigated, not individual loci.

Furthermore, if there is negative linkage disequilibrium of genes affecting the trait (i.e., plus factors linked with minus factors) then these chromosomal regions may be seen to have little effect. Therefore there is even more conservatism in the process.

There are always other genes of smaller effects than the limits of the resolution of the study.

This only tells you about which genes are responsible for the genetic differences between the two parental strains. Clearly, many others can be responsible for other variation.

Major methods
Single marker analysis- logistic regression

Lander and Botstein 1989

Regresses phenotype on presence and absence of markers

Simple, first published, but no longer the best way to do it

Interval mapping

Uses intervals, i.e. the length of chromosome between markers as the unit of analysis

Increasingly sophisticated: See papers and programs by Z.-B. Zeng

Recommendations

Use interval mapping
Use F2, not backcrosses
Use markers which are fixed in parental strains
Minimize variance within marker-classes

Multiple comparisons

This sort of analysis, like many others, requires that many different null hypotheses are tested on the same data set. therefore there will be many "false positives" by chance alone. This the significance level must be adjusted to a/n.

Maximum likelihood

Most analyses of QTL’s depend on ML methods. The log likelihood of a given mode is expressed as its LOD score (for log odds).

See figure 21.2 for an example, or almost any QTL paper (e.g. Long et al. 1995).

Experimental Results
There are generally multiple genes responsible for each trait, but not infinitely many of equally small effect

See Tables 21.4, 21.5

Additivity?

There is a large range of additive effects. See fig 21.3 for an example.

Dominance effects are also all over the map, ranging from strict additivity to total recessivity to overdominance.

Studies of mutations (e.g. Mukai) reveal that genes of large effect tend to have high dominance coefficients, while those of smaller effect are more likely additive. This is predicted by metabolic pathway theory.

Epistatic effects are found often, but not always. The current standard analysis is very weak for detecting epistatic effects.

QTL’s often have correlated effects

e.g. sternopleural and abdominal bristles in Drosophila.

Identifying QTL’s as genetic loci

Candidate loci are loci which we have an a priori reason to suspect are involved in the trait. If these loci fall within regions of high LOD scores, then they may be partially responsible for the quantitative variation attributable to that region.

This is tested by either complementation tests or by direct observation and experimental manipulation of the genotype at the candidate locus.