Figure 4

Feature probability profile construction for experimental data. Functional genomics experimental data are obtained associated with probes spanning and representing the entire genome at various resolutions (currently from some fifty base-pairs to hundreds of kilobases). The conversion of experimental data into probabilities requires the distinction of two cases. (A) In the trivial case only one experimental/biological condition has been investigated. In the absence of any additional information, the assumption that any probe returning a signal above its detection limit (DL) returns the maximal signal strength has to be made. Hence, any probe with a signal above its probe-specific threshold of detection will be assigned a probability of unity, whereas any probe-signal below the detection limit is set to zero. As discussed in the main text, all nucleotides covered by the probe are assigned the corresponding probability. For biological analysis this trivial case has little relevance. (B) For any thorough analysis several biological conditions C _q will be investigated using the same technology T. As discussed, the boundaries of the possible system states are unknown. With every new experimental dataset the probability layers for each condition need to be rescaled. For this, as biological data are generally assumed to follow lognormal distributions, we use the integral over a lognormal distribution as the rescaling function F^(T) _i . Obviously, any associated signal variance stemming from technical replicates is accordingly rescaled together with the probe probability.