Table 2 Results for 12 RNA targets in CASP15

From: trRosettaRNA: automated prediction of RNA 3D structure with transformer network

Target type

Target ID

RMSD (Å)

Yang-Server

AIchemy_RNA2

Chen

RNApolis

Deep learning besta

Overall best

Natural

R1107

17.9 (4.3b)

4.5

6.5

8.8

5.9

4.5

R1108

9.1 (4.8b)

4.5

6.0

8.5

4.8

4.5

R1116

10.9

17.3

18.0

12.7

7.9

4.8

R1117

2.7

2.3

2.0

2.7

2.7

2.0

R1149

13.9 (10.6b)

10.5

14.0

18.2

6.9

6.9

R1156

16.6

7.6

11.0

17.1

12.9

5.4

R1189

16.3

22.0

21.2

18.7

22.8

16.3

R1190

16.0

22.0

18.8

22.4

22.2

16.0

Average

12.9 (10.3b)

11.3

12.2

13.6

10.8

7.5

Synthetic

R1126

32.7

8.8

12.6

20.0

30.2

8.9

R1128

22.3

4.3

6.7

14.6

14.3

4.3

R1136

41.6

7.3

10.9

11.0

27.3

7.2

R1138

40.8

7.8

12.3

9.6

35.5

7.8

Average

34.4

7.0

10.6

13.8

26.8

7.0

Overall average

20.1 (18.3b)

9.9

11.2

13.7

16.1

7.4

  1. aAccording to the CASP15 abstracts, there are 14 RNA prediction groups utilizing deep learning-based methods to predict RNA structures.
  2. btrRosettaRNA results with secondary structure templates as inputs.
  3. For all compared groups, we evaluate their best-submitted models for each target. The evaluation based on the first predicted model is shown in Table S5.
Back to article page