Predicting Protein Cellular Localization Using a Domain Projection Method

Table 2.

Domains Whose Predicted Locales Differ from Their SMART Annotations

Domain Prob Reasonb
cytoplasmic → secreted
 4.1m 1.00 TM
 Calx_beta 0.99 TM
 TIR 0.99 TM
cytoplasmic → nuclear
 AAAa 0.59 I
 ARM 0.99 I
 BIR 0.83 I
 BTB 0.74 I
 CASc 0.76 I
 CARD 0.75 I
 MIR 0.95 I
 RING 0.72 I
 SPRY 0.88 I
 UBOX 0.73 I
 VHP 0.73 I
 ZnF_AN1 0.99 I
 ZnF_RBZ 0.85 I
 ZnF_UBPa 0.60 I
nuclear → cytoplasmic
 A1pp 0.59 I
 AT_hooka 0.53 (i)
 HLH 0.76 CI (PAS)
 HOX 0.77 CI (LIM)
 HSF 0.73 CI (REC)
 HTH_CRP 0.90 CI (cNMP)
 HTH_LUXR 0.70 CI (REC)
 MBT 0.95 I
 SFM 0.63 (ii)
 SMRa 0.64 I
 TOP2ca 0.60 CI (HATPase_c)
 TOP4ca 0.59 CI (HATPase_c)
nuclear → secreted
 Ku78 0.99 CI (VWA)
 SAND 0.91 (iii)
secreted → cytoplasmic
 BPI2 0.98 (iv)
 LysM 0.67 TM
 TSPc 0.94 CI (PDZ)
secreted → nuclear
 MATH 0.99 I
  • For example, cytoplasmic → nuclear means domains listed as cytoplasmic in SMART but predicted to be nuclear. Prob is the predicted locale probability of the domain. aDomain predicted as strongly multilocale. bTM: Domain occurs in transmembrane proteins. I: Indiscriminate domain for which there is literature evidence that the domain occurs in proteins found in more than one locale. CI: Domain is companion of an indiscriminate domain (listed). (i) AT_hook was wrongly predicted as cytoplasmic because of its close proximity in the domain projection plot to UBOX, an indiscriminate domain. (ii) SFM was wrongly designated as a nuclear domain in SMART. (iii) SAND was wrongly predicted as secreted because of an error in the domain architecture prediction by SMART of sequence Q9JLW9. (iv) BPI2 was predicted as nuclear rather than secreted as a result of a likely aberrant fusion with a PHD-containing sequence Q9LTR5.

This Article

  1. Genome Res. 12: 1168-1174

Preprint Server