What are "split reads" and "intron clusters?"

Question

I'm working through the example data set for LeafCutter and the documentation mentions "split reads":

This will cluster together the introns fond [sic] in the junc files listed in test_juncfiles.txt, requiring 50 split reads supporting each cluster and allowing introns of up to 500kb.

Searching "split reads" on Google yielded few results, and the results I did find were hard for me understand,

Chimeric reads occur when one sequencing read aligns to two distinct portions of the genome with little or no overlap. Chimeric reads are indicative of structural variation. Chimeric reads are also called split reads.

One sequencing read aligns to two distinct portions of the genome, but then with little or no overlap? Isn't that a self-contradiction?

I can more or less deduce what "intron clusters" are but I figured I might as well try to solicit a more formal definition. I know what introns are and I can figure what it means to cluster them together. I guess I just want to know what split reads and intron clusters mean in the context of working out the example data for LeafCutter.

Answer 1

split reads - These are read that have two or more alignments to the reference from unique region of the read. In this example a 150bp read sequenced from RNA could have base 1-75 aligning to the 3' end of exon2 and bases 76-150 aligning to the 5' end of exon3. This would be a split read because it have two alignments (exon2 and exon3) and those alignment are from unique (non overlapping) parts of the read, 1-75 and 76-150.

Chimeric reads are an example of split reads. Many people use "chimeric reads" to describe a type of noise that occur during PCR or sequencing where the bases from two different molecules are reported as one read and the alignments of that read become split when it aligns back to the location of the two original molecules.

"intron clusters" - In this case I think that the LeafCutter software is describing stacks of split reads to cluster to support the detection of an intron.