Hello, in experimenting with graph pruning options, I noticed after no more tips are removed, each subsequent cleaning round reports removing the identical number of unitigs. This is suspicious as it seems statistically unlikely that an arbitrarily large number of rounds in a row would decide to remove exactly the same number of unitigs, usually a number slighly less than the last round where >= 1 tip was pruned.
One example for k21 is shown below (although this behavior occurs at all k until much higher k > ~99 in most cases, after which it will early terminate before doing all rounds, as expected). Is this an error in reporting or is the program really disconnecting the exact same number of unitigs infinitely? Could it be trying to re-remove already-removed unitigs? Is there a risk of corrupting the DBG?
This happens in all metagenomes tested (25 of them), especially at lower k-mer sizes (< 99). In the log below, note where it says
"221 - Number unitigs disconnected: 397" 15 times in a row (or even more if a higher number of cleaning rounds is specified). Sometimes this repetition occurs as early as round 3 or 4, so even with the default number of cleaning rounds (5), this suspicious behavior occurs, but is less easy to notice.
Hello, in experimenting with graph pruning options, I noticed after no more tips are removed, each subsequent cleaning round reports removing the identical number of unitigs. This is suspicious as it seems statistically unlikely that an arbitrarily large number of rounds in a row would decide to remove exactly the same number of unitigs, usually a number slighly less than the last round where >= 1 tip was pruned.
One example for k21 is shown below (although this behavior occurs at all k until much higher k > ~99 in most cases, after which it will early terminate before doing all rounds, as expected). Is this an error in reporting or is the program really disconnecting the exact same number of unitigs infinitely? Could it be trying to re-remove already-removed unitigs? Is there a risk of corrupting the DBG?
This happens in all metagenomes tested (25 of them), especially at lower k-mer sizes (< 99). In the log below, note where it says "221 - Number unitigs disconnected: 397" 15 times in a row (or even more if a higher number of cleaning rounds is specified). Sometimes this repetition occurs as early as round 3 or 4, so even with the default number of cleaning rounds (5), this suspicious behavior occurs, but is less easy to notice.
This occurs with multiple datasets and at multiple k-mer lengths.