Wednesday, January 6, 2010

Penetrating the fog of 'penetrance'

In genetics there is a hoary old concept called 'penetrance'. It's not a definition of sexual success, so those of you who come to this post (no pun intended) with prurient interest should seek satisfaction elsewhere.

Penetrance is the probability that an individual has a specified trait given that s/he has a specified genotype. Usually, we think of the latter in terms of an allele, like the proverbial dominant A or recessive a in classical Mendelian terms in which genetics is taught.

Penetrance can range from zero -- the trait is never found in a person with genotype G -- to 1.0 (100% of the time the trait is present in persons with genotype G). If 'G' refers to an allele, that allele is called dominant if its presence is always associated with the trait, or recessive if the trait is present only in the absence of the other allele (in gg genotypes).

The key concept, that links simple Mendelian inheritance with general aspects of penetrance is that penetrance is almost always a relative term. The effect of an allele is always dependent on the other alleles in the individual's genome, as well as to aspects of the environment, and also to chance.

Sometimes things seem simple enough that we need not worry too much about these details. Very strong effects that are (almost) always manifest are examples. But when things are relativistic in this way, genetic inherency usually must take a back seat to a more comprehensive understanding.

The first step, and usually a difficult one, is to specify just what genotype you are referring to and, often even more challenging, just what phenotype (trait or aspects of a trait) you are referring to. To use the Einstein phrase that applies to relativity in physics, you have to be clear about your frame of reference.

This is much, much more easily said than done. Is 'heart disease' a useful frame of reference relative to alleles at some gene like, say, ApoE (associated with lipid transport in the blood)? We work with a colleagues, including Joan Richtsmeier here in our own department, who are concerned with craniofacial malformations. There are many such traits, including abnormal closure of cranial sutures (where bones meet in the skull). And, cancer is a single word that covers a multitude of syns (syndromes).

These are examples in which no two cases are identical. When that's so, how can we tell what the penetrance is of a mutation in a particular gene? Probabilistic statements require multiple observations, but also that each observation be properly classified (since probabilities refer to distinct classes of outcomes).

Since a given mutation affects only a single part of a single gene, it can be identified specifically (if, for the moment, we discount the mutations that take place within the person's body each time any of his/her cells divide). But traits can be variable and hard to define precisely, and the rest of the genome will vary in each person, even in inbred mice (because they undergo mutations). The amount of variation depends on the situation, but is very difficult to quantify precisely (recent work on genetic mutations in cancer begins to show this in detail, though we've known it in principle for a long time).

Most of the additional variation, not to mention purely chance aspects of development, homeostasis, and every cell's behavior, is unknown and much of it perhaps not even documentable in principle. Thus, in trying to characterize complex traits, we face real challenges just of definition, and much more so of understanding.

The same is true of evolution. An allele that has zero penetrance cannot be seen by natural selection. An allele with 100% penetrance is always 'visible' to selection in principle. But even there it has no necessary evolutionary implications unless it also affects fitness, that is, reproductive success. And that is another layer of causation with complex definitional issues, that we have written much about.

One bottom line is that just knowing a complete DNA sequence from some representative cell of an individual does not explain phenotypes, phenotypic effects, or evolution.

No comments: