About 'Complex non-cyclic'

[FromSoSimple]

Hi,

Thanks very much for sharing this tool to parse AmpliconArchitect (AA) results. My original understanding about AA's *cycles.txt file is that if a cycle include segment 0 at the start and at the end, then it's either linear amp or complex rearrangement; and if a cycle does not include segment 0, then it's cyclic: the last segment loops back to connect to the first segment.

But when I ran AmpliconClassifier (AC) on one of my AA outputs (amplicon1_cycle.txt file pasted below), where It's clear that Cycle 1, 2, 3, 4, 5 do not have segment 0, AC classified this amplicon as 'Complex non-cyclic'. I wonder why it's not classified as 'Cyclic'. Thank you!

Interval 1 X 127918918 127941545
Interval 2 X 128850387 129069878
Interval 3 X 131390129 131458996
Interval 4 X 134226913 134266980
Interval 5 X 135460501 135522871
Interval 6 X 142667656 142690042
List of cycle segments
Segment 1 X 127918918 127921210
Segment 2 X 127918918 127929347
Segment 3 X 127918918 127941545
Segment 4 X 127921117 127921210
Segment 5 X 127921117 127921296
Segment 6 X 127921151 127921210
Segment 7 X 128850387 128870947
Segment 8 X 128870383 128870983
Segment 9 X 128870383 128870985
Segment 10 X 128870433 128870697
Segment 11 X 128870433 128870826
Segment 12 X 128870433 128870844
Segment 13 X 128870433 128870947
Segment 14 X 128870433 128871516
Segment 15 X 128870443 128870682
Segment 16 X 128870443 128870947
Segment 17 X 128870474 128870826
Segment 18 X 128871077 128871423
Segment 19 X 128871077 128871441
Segment 20 X 128871077 128871490
Segment 21 X 128871077 128871497
Segment 22 X 128871077 128871545
Segment 23 X 128871178 128871423
Segment 24 X 128871178 128871490
Segment 25 X 128871178 128871566
Segment 26 X 128871178 128871920
Segment 27 X 128871210 128871490
Segment 28 X 128871210 128871516
Segment 29 X 128871210 129069878
Segment 30 X 128871321 128871516
Segment 31 X 128871397 128871545
Segment 32 X 128871780 129069878
Segment 33 X 131390129 131438835
Segment 34 X 131413108 131448398
Segment 35 X 131438260 131438835
Segment 36 X 131438310 131438436
Segment 37 X 131438310 131438835
Segment 38 X 131438434 131438436
Segment 39 X 131438434 131438850
Segment 40 X 131438521 131448398
Segment 41 X 131438541 131438683
Segment 42 X 131438541 131438835
Segment 43 X 131438568 131438683
Segment 44 X 131447145 131448398
Segment 45 X 131452790 131456186
Segment 46 X 131452790 131458996
Segment 47 X 134226913 134247009
Segment 48 X 134233109 134233361
Segment 49 X 134246839 134247009
Segment 50 X 134246854 134246980
Segment 51 X 134246854 134266980
Segment 52 X 134246913 134247009
Segment 53 X 135460501 135522871
Segment 54 X 135480390 135503158
Segment 55 X 135502443 135502601
Segment 56 X 142667656 142690042
Segment 57 X 142689572 142690042
Cycle=1;Copy_count=10.7116740972;Segments=9+,5-
Cycle=2;Copy_count=7.13433873533;Segments=37+,25+,17+
Cycle=3;Copy_count=4.3826755996;Segments=39+,15-
Cycle=4;Copy_count=2.3093359841;Segments=12+,19-
Cycle=5;Copy_count=2.21980533822;Segments=13+,13-,21+,6+,20-
Cycle=6;Copy_count=2.19718248634;Segments=0+,56-,0-
Cycle=7;Copy_count=2.09796638102;Segments=8+,49+,27+,4-
Cycle=8;Copy_count=2.03240257232;Segments=26+,35+
Cycle=9;Copy_count=1.98836332533;Segments=0+,33+,24+,1-,0-
Cycle=10;Copy_count=1.14207766849;Segments=0+,3-,0-
Cycle=11;Copy_count=1.06159424616;Segments=0+,53-,0-
Cycle=12;Copy_count=1.01079937939;Segments=0+,29-,47-,0-
Cycle=13;Copy_count=0.951940523565;Segments=0+,51-,38-,16+,7-,0-
Cycle=14;Copy_count=0.677046785145;Segments=42+,23+
Cycle=15;Copy_count=0.550129044225;Segments=34+,45+
Cycle=16;Copy_count=0.526777056708;Segments=54+,14-,22+
Cycle=17;Copy_count=0.503671349318;Segments=0+,46-,44-,48+,32+,0-
Cycle=18;Copy_count=0.405645559019;Segments=54+,28-,52-,31+
Cycle=19;Copy_count=0.309049994232;Segments=0+,2+,43-,55+,50-,36-,11-,18+,41+,2-,0-
Cycle=20;Copy_count=0.250119950562;Segments=54+,30-,10-,22+
Cycle=21;Copy_count=0.249063483838;Segments=0+,46-,40-,57+,0-

[FromSoSimple]

Looks like their lengths might be too short?

[jluebeck]

Hi,

Sorry for the delay in responding. The classifier by default expects the cycles which comprise the ecDNA will be at least 10 kbp on each cyclic path. At present, we do not attempt to make determinations about the origins of cyclic decompositions which are significantly smaller than that.

Best,
Jens

[FromSoSimple]

Thanks, Jens!

[FromSoSimple]

Hi Jens,

I have another case with following AA results:

Interval 1 2 15662961 16718671
List of cycle segments
Segment 1 2 15762954 15982958
Segment 2 2 15762954 16271651
Segment 3 2 15984443 16108827
Segment 4 2 15984443 16120775
Segment 5 2 16109441 16120775
Segment 6 2 16123800 16271651
Segment 7 2 16273865 16282960
Segment 8 2 16332961 16402960
Segment 9 2 16332961 16618682
Segment 10 2 16422961 16484159
Cycle=1;Copy_count=129.66200028;Segments=0+,7-,2-,9-,0-
Cycle=2;Copy_count=70.7033566593;Segments=0+,7-,6-,5+,3+,1-,9-,0-
Cycle=3;Copy_count=18.1865904689;Segments=4+
Cycle=4;Copy_count=17.1845928766;Segments=0+,8-,0-
Cycle=5;Copy_count=12.0828255791;Segments=0+,10-,0-

To me Cycle 3 is cyclic according to AA manual, but AC classifies this case as Complex non-cyclic. Any idea why?

[jluebeck]

Hi,

The decompositions produced by AA are not classifications of the amplicon into a biological class of amplification. They represent a more abstract decomposition of the breakpoint graph structure. The presence of cyclic graph decompositions is indeed associated with ecDNA. However, cycles in the graph can arise by different mechanisms, such as BFB. You can see the more biological classifications in the AC output columns which indicate ecDNA+ or BFB+.

In this case, it appears the vast majority of the amplicon CN is explained in non-cyclic structures. Thus, it is likely AC does not have sufficient evidence to bioinformatically call an ecDNA in this case. Due to the rearranged profile of the segments, it then applies the "complex non-cyclic" amplicon classification. Note that AC having insufficient evidence to call an ecDNA does not imply the sample is certainly negative for ecDNA.

Best,
Jens

[FromSoSimple]

Got it. Thanks very much, Jens.