Table 7

The probabilistic approach of PFold can, on occasion, suffer from "under-flow" errors caused by multiplying many probabilities together producing numbers too low to be dealt with on modern computers. This is what has happened on the medium similarity data-set.

E. coli SSU rRNA: Comparative Methods


Algorithm

number of bps in reference

number of bps in prediction

True Positives (% sensitivity)

False Positives (% selectivity)

Correlation (%)

Plan A: ClustalW Alignment

RNAalifold (H)

460

472

275 (59.8)

179 (60.6)

0.601 (60.2)

RNAalifold (H) + RNAfold-C

460

483

273 (59.3)

195 (58.3)

0.588 (58.8)

RNAalifold (M)

441

433

372 (84.4)

32 (92.1)

0.881 (88.2)

RNAalifold (M) + RNAfold-C

441

469

388 (88.0)

44 (89.8)

0.889 (88.9)

Pfold (H)

460

377

326 (70.9)

26 (92.6)

0.810 (81.7)

Pfold (M)

441

0

0 (0.0)

0 (0.0)

0.000 (0.0)

ILM (H)

460

565

236 (51.3)

313 (43.0)

0.469 (47.1)

ILM (M)

441

564

264 (59.9)

249 (51.5)

0.554 (55.7)

ILM-pknot (H)

468

565

236 (50.4)

311 (43.1)

0.466 (46.8)

ILM-pknot (M)

468

564

266 (56.8)

258 (50.8)

0.537 (53.8)

Plan B: Unaligned sequences

Carnac (H)

460

233

206 (44.8)

12 (94.5)

0.650 (69.6)

Carnac (H) + RNAfold-C

460

470

332 (72.2)

112 (74.8)

0.734 (73.5)

Carnac (M)

448

294

259 (57.8)

18 (93.5)

0.735 (75.7)

Carnac (M) + RNAfold-C

448

471

337 (75.2)

110 (75.4)

0.753 (75.3)


Gardner and Giegerich BMC Bioinformatics 2004 5:140   doi:10.1186/1471-2105-5-140

Open Data