Table 2 Average error rate of seven methods in inferring order of mutations.

From: Power and pitfalls of computational methods for inferring clone phylogenies and mutation orders from bulk sequencing data

Branch*

Method

CloneFinder

MACHINA

TreeOmics

LICHeE

MixPhy

PhyloWGS

Cloe

G7 dataset

Concurrent

2% (4%)

4% (5%)

10% (4%)

20% (11%)

10% (4%)

14% (11%)

0% (0%)

Sequential

4% (5%)

7% (5%)

5% (3%)

17% (10%)

9% (2%)

16% (12%)

2% (6%)

Parallel

3% (5%)

5% (4%)

4% (4%)

19% (17%)

9% (4%)

25% (14%)

2% (6%)

G12 dataset

Concurrent

1% (2%)

5% (3%)

12% (5%)

0% (1%)

0% (1%)

27% (7%)

14% (2%)

Sequential

1% (2%)

7% (4%)

6% (3%)

5% (1%)

38% (2%)

37% (6%)

26% (11%)

Parallel

0% (1%)

4% (2%)

4% (2%)

3% (1%)

15% (2%)

34% (12%)

10% (3%)

P10 dataset

Concurrent

11% (7%)

13% (9%)

13% (9%)

10% (7%)

11% (7%)

13% (7%)

10% (6%)

Sequential

10% (10%)

11% (10%)

11% (11%)

7% (6%)

17% (17%)

19% (13%)

14% (9%)

Parallel

3% (5%)

3% (6%)

1% (4%)

1% (2%)

6% (10%)

11% (13%)

5% (7%)

MA dataset

Concurrent

19% (7%)

13% (8%)

20% (8%)

16% (8%)

19% (7%)

21% (7%)

NA

Sequential

18% (9%)

11% (7%)

18% (8%)

20% (13%)

34% (19%)

27% (12%)

NA

Parallel

5% (5%)

4% (4%)

7% (5%)

5% (5%)

8% (6%)

12% (8%)

NA

Average

6%

7%

9%

10%

15%

21%

NA

  1. Standard deviation is shown in a aprenthesis.
  2. *We generated all possible pairs of mutations (SNVs) and tested if each pair of mutations were correctly ordered along mutational tree. There are three possible orders of mutations, (1) one mutation is placed at ancestral or descendant branch of the other metation (Sequential); (2) mutations are placed at different lineages (Parallel); (3) mutations are placed at a same branch (Concurrent).
  3. We measured the accuracy by computing the error rate, which was the average of the proportion of correct orders not found and that of incorrect orders produced. Smaller values are better. Methods were sorted by the overall average across datasets and branche categories.