As the Fig.1a in StringTie paper (https://www.nature.com/articles/nbt.3122), we know StringTie extracts the heaviest path as a transcript from the alternative splice graph. And the heaviest path starts at the node with the highest reads per base coverage, then extends the path first to the source and then to the sink by choosing the adjacent node with the highest number of compatible fragments with the path built so far. According to the paper, two fragments are considered to be compatible if one fragment starts in the same node as the other one ends.
But I still have some questions about this description.
1.If there are two reads mapped with partial overlap, and they are in the same node. Are they compatible? I think they follow the rule “one fragment starts in the same node as the other one ends”.
2.How to calculate the compatible number if there are more than two fragments? From my understanding, in this case, fragment A is compatible with C. And fragment A is also compatible with D, E, F, G. What is the accurate compatible fragments number in this case?
3.How to calculate the compatible fragments between two nodes? From my understanding, in this case, fragment C is compatible with A and E. How many compatible fragments are there between node1 and node2?
If my examples are not appropriate, please tell me. I wonder the clear definition about “compatible fragment” in StringTie, and how to calculate the precise number.