Sciclone took too long

Question

[sm.hashemin]

Dear Chris,

I was trying sciclone on my data and it took one day and it was

    [1] "checking input data..."
    [1] "Not all variants fall within a provided copy number region. The copy number of these variants is assumed to be 2."
    [1] "Not all variants fall within a provided copy number region. The copy number of these variants is assumed to be 2."
    10284 sites (of 19026 original sites) are copy number neutral and have adequate depth in all samples
    238 sites (of 19026 original sites) were removed because of copy-number alterations
    8742 sites (of 19026 original sites) were removed because of inadequate depth
    8742 sites (of 19026 original sites) were removed because of copy-number alterations or inadequate depth
    [1] "clustering..."

Disable overlapping std dev condition
kmeans initialization:
Tumor.vaf   Relapse.vaf
0.554656036649214   0.0234683628795812
0.361785698504027   0.423195796317607
0.0176328924516819  0.0381666242562692
0.579921962920046   0.525048517960602
0.30080528290469    0.0624659469531013
0.392210176923077   0.991532659340647
0.0132568691148545  0.472080169984686
0.997534324719097   0.996064462921345
0.995766899999987   0.434549282777778
0.118433075498615   0.0728657581502769
Using threshold:  0.83666

I stopped the process here, and then I got these messages:

Warning messages:
1: did not converge in 10 iterations 
2: Quick-TRANSfer stage steps exceeded maximum (= 514200) 
3: Quick-TRANSfer stage steps exceeded maximum (= 514200)

could I some how avoid this? Best Mo

[GenoMax]

Tagging: Chris Miller

[sm.hashemin]

Hi there! It is indeed a WES at 200X, I did a tumor-only variant calling with LoFreq. Mutect2 tumor-only gives me even more variants. I tried puting stringent filters like depth of >200, VAF >10%, SB>30. The number was not significantly reduced! What would you recommend for variant calling? should I go for Pool of healthy individuals to call somatic only? I can not get blood samples .

Best Mo

[GenoMax]

Please use ADD COMMENT/ADD REPLY when responding to existing posts to keep threads logically organized.

This comment belongs under @Chris' answer.

[Chris Miller]

Tumor only means that you will almost certainly not be able to remove all germline mutations. The isown paper will provide some useful guidelines for how you can minimize them though. https://genomemedicine.biomedcentral.com/articles/10.1186/s13073-017-0446-9

[2891913164]

Dear Chris,

I also encountered the same problem. I extracted the relevant information needed by sciclone from the vcf file generated from the paired normal tumor data as input, and then the following problem has been encountered. I would like to ask whether a certain screening should be carried out when extracting data from the vcf file? If I need to filter the vcf file, what criteria should I use to filter it? Hope to get your help, thank you very much!

[1] "checking input data..."
12462 sites (of 12480 original sites) are copy number neutral and have adequate depth in all samples
0 sites (of 12480 original sites) were removed because of copy-number alterations
18 sites (of 12480 original sites) were removed because of inadequate depth
18 sites (of 12480 original sites) were removed because of copy-number alterations or inadequate depth
[1] "clustering..."
Disable overlapping std dev condition
kmeans initialization:
V1
0.369202574615969
0.0451805969314854
0.0012764317159691
0.068427701248344
0.130563008213922
0.508433685628903
0.419371045057848
0.567150829909654
0.0267695356487535
0.23147262643314
0.998687595233288
0.0133681853132312
0.00484191691530496
0.096954918736847
0.767968455466162
0.306889133938259
0.893953674722921
0.171738804853286
0.647833820706109
0.469620180355922
Using threshold:  0.7 
Dropped cluster 5 with too few variants ( 0 ) center: 0.4961276
Dropped cluster 5 with too few variants ( 0 ) center: 0.4889433
Dropped cluster 5 with too few variants ( 0 ) center: 0.4961276
Dropped cluster 7 with too few variants ( 0 ) center: 0.4961276
Dropped cluster 11 with too few variants ( 0 ) center: 0.4961276
Dropped cluster 12 with too few variants ( 0 ) center: 0.4961276
Dropped cluster 12 with too few variants ( 0 ) center: 0.228628
Cluster 1 pi = 0.381 center = 0.382 SEM = (0.372, 0.393) sd = (0.000, 1.000)
Cluster 2 pi = 0.557 center = 0.004 SEM = (0.004, 0.004) sd = (0.000, 0.006)
Cluster 3 pi = 0.062 center = 0.479 SEM = (0.476, 0.482) sd = (0.417, 0.544)
Converged on the following parameters:
mu:
3088.34643978221 88967.680080543 35087.7594185802
alpha:
57064.3763044592 40966.6875698716 574.763038704414
nu:
1720.59558497192 15169.4708677206 33819.9881156314
beta:
19643.4195415117 25.601254166207 510.134678204943
pi:
0.380504889742695   0.557239249326397   0.0622558639309075
[1] "finished clustering full-dimensional data..."
[1] "found 3 clusters using bmm in full dimensional data"
There were 50 or more warnings (use warnings() to see the first 50)
> writeClusterTable(sc, "test.new/clusters1")
> sc.plot1d(sc,"test.new/clusters1.1d.pdf")
> 
 warnings()
1: did not converge in 10 iterations
2: did not converge in 10 iterations
3: did not converge in 10 iterations
4: did not converge in 10 iterations
5: did not converge in 10 iterations
6: Quick-TRANSfer stage steps exceeded maximum (= 623100)
7: did not converge in 10 iterations
8: did not converge in 10 iterations
9: did not converge in 10 iterations
10: did not converge in 10 iterations
11: did not converge in 10 iterations
12: did not converge in 10 iterations
13: did not converge in 10 iterations
14: did not converge in 10 iterations
15: Quick-TRANSfer stage steps exceeded maximum (= 623100)
16: did not converge in 10 iterations
17: did not converge in 10 iterations
18: Quick-TRANSfer stage steps exceeded maximum (= 623100)
19: Quick-TRANSfer stage steps exceeded maximum (= 623100)
20: Quick-TRANSfer stage steps exceeded maximum (= 623100)
21: did not converge in 10 iterations
22: Quick-TRANSfer stage steps exceeded maximum (= 623100)
23: did not converge in 10 iterations
24: Quick-TRANSfer stage steps exceeded maximum (= 623100)
25: did not converge in 10 iterations
26: did not converge in 10 iterations
27: did not converge in 10 iterations
28: Quick-TRANSfer stage steps exceeded maximum (= 623100)
29: did not converge in 10 iterations
30: did not converge in 10 iterations
31: did not converge in 10 iterations
32: did not converge in 10 iterations
33: did not converge in 10 iterations
34: did not converge in 10 iterations
35: did not converge in 10 iterations
36: did not converge in 10 iterations
37: did not converge in 10 iterations
38: did not converge in 10 iterations
39: did not converge in 10 iterations
40: did not converge in 10 iterations
41: did not converge in 10 iterations
42: did not converge in 10 iterations
43: did not converge in 10 iterations
44: did not converge in 10 iterations
45: Quick-TRANSfer stage steps exceeded maximum (= 623100)
46: did not converge in 10 iterations
47: did not converge in 10 iterations
48: did not converge in 10 iterations
49: did not converge in 10 iterations
50: did not converge in 10 iterations
> 

Answer 1

[Chris Miller]

Are your 19k sites from exome or WGS? If the former, I suspect that something has gone horribly wrong with your mutation calling. In a decade of working on tumors, I've only seen a handful of MMR/POLE deficient cases with that kind of mutational burden in an exome.

Sciclone will also take a long time if your data is very noisy. Do a simple X/Y plot of your VAF data and see if you can make out the clusters by eye. If not, then the algorithm is likely going to have a hard time too. You might consider being more stringent with your variant calling parameters and requiring a higher minimum depth to tighten things up.