confusion in read assignment

[bluemood2011]

hi,

I am using isoquant with my pacbio isoseq data and confused by the read assignment strategy. Here is an example:
In the file OUT.read_assignments.tsv.gz, one read is assigned to two transcripts, one is fsm and the other one is ism:
m84128_251204_111751_s4/74582568/ccs/980_2261 NC_000011.10 + NM_003977.4 9049 ambiguous fsm,tss_match_precise:4,tes_match_precise:-2,correct_polya_site_right:67491102 67483022-67483257,67487006-67487185,67489267-67489455,67490038-67490214,67490316-67490457,67490788-67491101 gene_assignment=unique; PolyA=True; Classification=full_splice_match;
m84128_251204_111751_s4/74582568/ccs/980_2261 NC_000011.10 + XM_024448761.1 9049 ambiguous ism_5,tes_match_precise:-2,correct_polya_site_right:67491102 67483022-67483257,67487006-67487185,67489267-67489455,67490038-67490214,67490316-67490457,67490788-67491101 gene_assignment=unique; PolyA=True; Classification=incomplete_splice_match;

I set the --transcript_quantification all, which will assign the ambiguous reads equally to each transcript as described in the documentation, but in the file OUT.transcript_model_reads.tsv.gz, the read is only assigned to XM_024448761.1.

[andrewprzh]

Hi @bluemood2011

The only explanation I see is that one isoform is a subsequence of another.

Something like:

Read                              [     ]-------[      ]--------[      ]AAAAA
NM_003977.4                       [     ]-------[      ]--------[      ]
XM_024448761.1           [   ]----[     ]-------[      ]--------[      ]

Since there is no indicator that the read is complete on 5' (e.g. similar to polyA tail on 3'), IsoQuant assumes the read might be truncated and there is no sure way to tell which isoform is the original one.

As more users request this feature, I will soon implement a "full-length" matching strategy where all reads are assumed to be complete on both sides.

Best
Andrey