Figure 3.

Summary of Gentrepid predictions for the seeded and ab initio modes of the CPS and CMP modules averaged across phenotypes. For each SNP-to-gene approach, the average proportion of the total predictions returned by CPS (represented as squares), CMP (represented as triangles), and the intersecting results from both modules (represented as diamonds with joining lines) are shown. The proportion of predictions refers to the ratio of the average number of genes predicted by each module to the average number of predictions made by both modules (i.e. CPS ∩ CMP). The four SNP sets employed are shaded yellow for HS, orange for MHS, purple for MWS and dark purple for WS. The proportion of predictions made by CPS in seeded mode was generally larger than the proportion predicted by CPS in ab initio mode. In ab initio mode, CMP makes the lion's share of predictions in search spaces with an intermediate number of annotated genes. For example, for the WS set which has the largest number of pseudo-intervals, the average size of the largest search space is 2285 annotated genes. CMP makes proportionally more predictions than CPS for the four smaller pseudo-intervals (total average gene set sizes of 155-803 annotated genes). For several of the HS sets no predictions were made and hence are superimposed on the x-axis. The average number of predictions per mode, module and WTCCC SNP set is shown in Table 2.

Ballouz et al. BMC Genetics 2011 12:98   doi:10.1186/1471-2156-12-98
Download authors' original image