Statistical Analysis of Genetic Syndromes
Because of computer technologies, genetics has benefited during the past 20 years from specific and complex explanations of the mechanisms leading to genetic syndromes and diseases. Genetic syndromes are complex occurrences with genotype aberrations, with or without observable (phenotypic) characteristics, that can seriously affect the normal life of individuals. Embryological and fetal anomalies (abnormalities) that may or may not include genetic defects, are classified as individual anomalies (e.g., malformation, disruption, deformation, dysplasia), and pattern anomalies (e.g., associations, sequences, syndromes, and developmental field defects). With computers, statistical analyses of such diseases can be accomplished easily, including evaluation of the frequency of a single malformation in the presence of a genetic disease, or in complex processes such as genetic transmission analysis (linkage studies), and multivariate analyses of patterns and risk factors.
Linkage is the occurrence of two or more genes (genetic loci) with a higher probability to segregate together rather than independently during meiosis. Linkage occurs because crossing over does not usually take place between loci that are close to each other. The unit used to express how close two linked genes are is the centimorgan (cM), or percent recombination. The statistical method of measuring linkage is the logarithm of the odds (Lod score) that expresses the odds in favor of finding the observed combination of alleles at the loci being studied if they are linked at the given distance, rather than being unlinked. A specific cut-off point of Lod score of +3 is considered strong evidence of linkage. Linkage allows the determination of the likely genotype of an individual, and the patterns of inheritance of a specific form of a heterogenetic disease. Linkage is also useful in determining the role of genetic factors in heterogeneous conditions, as well as cleft lip and palate and insulin-dependent diabetes. Clinical uses of linkage include prenatal diagnosis, carrier detection and pre-symptomatic diagnosis.
Cluster analysis is a wide set of multivariate techniques that attempts to identify relatively homogeneous groups of cases (or variables) based on selected characteristics, as well as phenotype patterns, using an algorithm that starts with each case (or variable) in a separate cluster and combines clusters until only one is left. In the analysis of genetic syndromes, cluster analysis can be useful in finding new syndromes based upon a recurrent pattern of phenotypic characteristics. Cluster analysis also yields the accuracy of the statistical model to properly classify the subjects, if the dependent variable (the ascertained genetic disease, for example) is known. A new application of cluster analysis is represented by the microarray. This technology promises to monitor, by means of sophisticated software, the whole genome on a single chip so that researchers have a better picture of the simultaneous interactions among thousands of genes. A stepwise selection of gene expression that has a different pattern in pathologic cases when compared to controls (experimental standards) allows a proper prediction model able to swiftly classify the subjects.
Latent class analysis is another sophisticated model that can be used for assessing the validity of discrete measurement, such as categorical results of some non-definitive diagnostic criteria. Initially, latent class analysis was mainly used for studies in psychology. In this statistical approach, the observed data (e.g., phenotype anomalies) are considered indicators of a non-directly observable variable for a genetic disease. Therefore, different realizations correspond to different patterns of observations. Latent class analysis is able to identify these latent types of patterns that can also be used to estimate the conditional probability for a subject belonging to the assigned class. Latent class analysis is also used to evaluate how a pattern of malformations can correctly detect a specific phenotype associated with a karyotype abnormality.
Non-invasive prenatal genetic research is a growing field that has as its main aim the selection of a population of pregnant women with a higher risk of carrying a fetus affected by genetic disease. One statistical method used is the Bart test, which calculates the probability of the diseases using quantitative biochemical markers (or a sonographic marker called nuchal translucency) able to adjust the age-specific probability of a syndrome for an appropriate correction factor for the individual called the likelihood ratio. The likelihood ratio is approximately homogenous with maternal age probability, and their combination yields an adjusted and more correct probability of the disease (also called odds of being affected given a positive result, or OAPR). An analogue of the Bart test for qualitative markers (as well as sonographic findings) is logistic regression, that evaluates -- by means of a slightly different factor of correction named the odds ratio -- the association of a positive marker (also called risk factor) to the pathology (also called outcome). The odds ratio is a very useful method to understand the magnitude of an association between a qualitative variable with an event. It is expressed as the ratio between the risk of observing a disease with and without a risk factor. The logistic regression output is usually a bit less reliable than the Bart test, and it must, therefore, be used and evaluated with care.
Several new models are available for the evaluation of genetic diseases. The use of such techniques is strictly relegated to their clinical applicability, and only after after careful patient counselling. In fact, because of the relative rarity of genetic syndromes and the enormous span of phenotype patterns, it is difficult to collect a population to produce "robust" results unaffected by bias and inaccurate predictive values.
This is the complete article, containing 900 words
(approx. 3 pages at 300 words per page).
