An enduring medical need is the better understanding of the determinants of human traits such as disease, or response to drugs. Genes and their protein products are the inherited building blocks of metabolic pathways that drive human physiology. The advent of high speed, high density genotyping offers a new opportunity to probe such. Binary single nucleotide polymorphisms (SNPs) at millions of nuclear chromosomal loci can now be employed to genetically characterise trait cases versus non-trait controls.

Fundamental to this aspiration is the classification of people, in particular how much evidence an individual has that indicates that they are genetically distinct from another. This is difficult as humans are 99.9% identical at the DNA level, with the possibility that only one SNP change in three thousand million bases may cause a trait of interest. With any two people having approximately one million SNPs different than each other, a simple, sensitive, easy-to-calculate but unified way of initially handling this taxonomic question for individuals is needed.

Large volume, whole genome SNP scans of humans are difficult to deal with. There are published methods that ascribe individuals to evolutionary groups using genetic data. However, although a variety of multi-locus approaches exist for marker mapping, few widely accepted established methods, bar small-scale haplotype estimation and their comparison, exist to attack the simultaneous processing of the legions of polymorphisms in whole genome scans. Researchers rely on 'one at a time' (univariate) tests beset with multiple testing problems. Some combinatorial and partitioning methods show promise as does multifactor dimensionality reduction and symbolic discriminant analysis. However, many potentially widely touted multivariate methods such as artificial neural networks or support vector machines produce impenetrable 'black box' solutions. Possible other methods for consideration include cellular automata and evolutionary computation.

All the above methods can appear highly complex to a lay reader and are certainly computer intensive, often requiring highly specialised staff to achieve. We believe that, what is needed is a straightforward method that enables ordinary scientists and clinicians themselves to expose useful biological and medical insights from whole genome scan SNP association studies.

top of page