How to infer the TSS from a gtf file

I have a csv file that I’ve downloaded from this paper here (abridged version below), and it’s referring to CpG locations in the genome from beadChip array data, using assembly hg18 as follows:

CpGmarker   Build   Chr MapInfo SourceVersion   TSS_Coordinate  Gene_Strand Symbol  Synonym Accession   GID
cg00075967  36  15  72282407    36.1    72282245    -   STRA6   PP14296; FLJ12541;  NM_022369.2 GI:21314699
cg00374717  36  17  63814740    36.1    63815191    +   ARSG    KIAA1001;   NM_014960.2 GI:45430056
cg00864867  36  12  78609399    36.1    78608921    -   PAWR    PAR4; Par-4;    NM_002583.2 GI:55769532
...

Let’s focus on the first entry for now, STRA6. I downloaded the Gencode gtf files, and the corresponding lines look like this (again, abridged for simplicity):

#bin    name            chrom   strand  txStart         txEnd           cdsStart    cdsEnd  exonCount   exonStarts  cdsEndStat  exonFrames  
1136    ENST00000395105 chr15   -   72258859    72282273    72259473    72281661    19  72258859,72260175,72260688,72261554,72261736,   STRA6   cmpl    cmpl
1136    ENST00000323940 chr15   -   72258859    72288424    72259473    72281661    19  72258859,72260175,72260688,72261554,72261736,   STRA6   cmpl    cmpl

The general locations line up pretty closely with the csv above, but am I being naive for thinking that the “TSS” should just be the furthest upstream position in the gtf? i.e. the lowest position within an exon (or the txStart) for genes on the + strand (highest for “-” strand ) ?
It seems obvious, but I’m doubting myself because none of these values agree with the value that is provided for the “TSS” in the csv of the publication. I’ve also tried with RefSeq and a few others –nothing seems to match up precisely.

So am I reading the gtf files correctly to infer the TSS (is it just txStart for “+” strand and txEnd for “-“?), and if so, is there some other obvious reason why the TSS coordinates in the paper’s csv file aren’t the same?

Thanks for any help you can offer.

Read more here: Source link