Bin Yu

2. INFORMATION THEORY and STATISTICS

2.2 Clone Mapping of Chromosomes in Human Genome Project

The primary goal of the Human Genome Project is to sequence the entire human genome, which consists of about $3 \times 10^9$ base pairs (bp) of DNA. Current technology permits sequencing of fragments of the order of only a few hundred to one thousand base pairs of DNA in a single reaction. Consequently much effort is devoted to fragmenting large DNA molecules such as chromosomes in such a way that the sequenced fragments, called clones, can be readily assembled. Clone maps, a form of physical mapping, play a key role in this process and allow the detailed study of chromosomal regions of biological interest. Formally, a clone map of part or all of a chromosome is constructed from information on order and overlap extracted from collections of DNA fragments called clone libraries. When comparing the relative merits of fingerprinting cloned fragments of DNA by competing molecular biology techniques (such as random fingerprinting methods, which include hybridization and restriction digestion), biologist H. Lehrach raised two interesting questions concerning the creation of clone maps information is gained by the hybridization and restriction digestion methods, respectively? answers Lehrach's first question by formulating a number of different notions (or configuration variables) of a clone map. The entropy of each notion (or configuration variable) is either calculated or bounded using combinatorics, Jensen's inequality and entropy equalities and inequalities. The bounds are derived under the same theoretical model on random clone libraries as in the works of Waterman and coauthors. The implications of our results are similar to theirs in that both can be useful for planning future mapping efforts. In particular, based on the bounds in Yu and Speed (1996), comparisons are made for four ``model" species in terms of information needed for the mapping of their respective cosmid clone libraries. It follows that the cosmid clone mapping for the roundworm requires about 40 times as much information as that for the bacterium E. coli, and that such mapping for humans requires about 1,500 times as much information as that for the bacterium E. coli. Another interesting fact which follows from the entropy bounds is that a variable relating to the pairwise approach to clone mapping contains a substantial proportion (more than 20\%) of the information in a full configuration variable. Various random fingerprinting methods are sometimes used to detect the overlap between pairs of clones as a first step towards producing a minimal tiling path of clones for subsequent mapping and sequencing efforts. Research is under way that uses the results in Yu and Speed (1995) to answer Lehrach's second question on how much information one can learn about a clone map using data collection techniques such as random fingerprinting methods. analyzes the overlap detection problem for two clones. We evaluate and compare various statistical procedures for the two-clone overlap based on random fingerprinting data. In particular, we quantify the limitations of random fingerprinting as a way to detect pairwise overlap, and within those limitations, the most effective ways to use the data. Based on our results, we conclude that random fingerprinting based methods generate very weak overlap detectors, confirming what biologists are discovering by trial and error. Moreover, although the best way to use the data is to employ a Bayes detector, well designed linear detectors can perform nearly as well as a more complex Bayes detector. It also follows from the results that measurement errors quickly impair the power of even well designed detectors, and if one wishes to use a Bayes detector, accurate analytical approximations and table look up methods are easier and quicker to use than numerical approximations.