Figure 7.
Output comparison of the annotation-supported model and the ARTS TSS prediction program. Chromosome 16 is divided into 50-nt chunks, and the prediction having the largest value within each chunk is computed for each program. Positive chunks contain RefSeq or CAGE starts, respectively. Negative chunks comprise downstream gene portions for RefSeq, and all non-positive chunks for CAGE. The area under the curve (auROC) provides a performance measure given these chunk definitions.
