Main Results

Results presented in this section are for the base scenario, which has the following parameters:

• Length of clones: l_c = 40960 base pairs.
• Target genome length: L = 25 clone lengths ( $L = 25 \times 40960$ which is approximately 1M base pairs).
• Clone coverage: 10
• False negatives probability:
• False positives probability: $\beta = 0.05$
• Number of probes: n = 500
• Length of probes: plen = 8

Based on the results of 1000 simulations, the algorithm has a probability of $0.039 \pm 0.006$ of making any big errors in the base scenario. The average error in the constructed map is also quite small: $1740 \pm 3 bp.$ The average error can be reduced if a finer quantization unit is used (at a linear cost to memory and CPU consumption). A further experiment indicated a probability of about 0.075 of making any mistake that exceeds a clone's length in the estimation of the relative distance between any two (not necessarily adjacent) clones.

  

Figure: True clone order (y axis) vs. constructed order (x axis) in four scenarios. When applicable, weak points shown as vertical dotted lines. The results are taken from the following scenarios: (a) base scenario, (b) a long 2MB genome, (c) a simulation with very low coverage (5), (d) a simulation with very low coverage (5) and very high noise ( $\alpha = 0.3$ and $\beta = 0.25$). Note that all big errors were detected as weak points, though some weak points incorrectly suggested additional big errors. pinpoint possible errors. This information can be used for a judicious choice of additional hybridization experiments, minimizing cost and human effort.

  

Figure 9.13: Influence of various simulation parameters on the probability of having big errors. The vertical dotted line indicates the value of the parameter in the base scenario. Note that the effect of a decrease in the number of probes is very similar to that of an increase in the experimental noise. This is because noise decreases the informational content of each probe, an effect that can be countered by an increase in the number of probes. It is also notable that the probe size has a very significant effect, resulting from its direct influence on the frequency of probe occurrences, and therefore on the informational content of the experiment. In contrast, the genome size only moderately effects performance.

 

 

In analogy to the breaking ­ up of a chemical molecule, the separation of a contig into two non­overlapping parts should increase the energy substantially. However, if the two parts do not overlap in the real map, the separation energy should be quite small or even negative.  

Such information can be used by the laboratory in order to pinpoint areas where additional hybridizations should be performed. We also make use of the weak points in our algorithm in order to break up a contig and reassemble it. In case an error was made at an early stage, this process enables the algorithm to correct its previous error with the benefit of the additional information from other clones added at a later stage. An example of weakpoint detection is described in Figure 9.12.