Protocol Online logo
Top : Forum Archives: : Bioinformatics and Biostatistics

Tss prediction - help (Apr/17/2007 )

Hi, can anyone give me a hand with this...

I need to find TSS for a gene, I 've alredy read a paper in which they found the TSS in silico, using Core promoter program. I read the original paper "Identification of human gene core promoters in silico" but I don't understand the whole thing and after doing the analisis for one of the sequences in the paper, I don't get the same results so I must be doing something wrong.

Can anyone help me? I can used other programs but, if I don't do an experimental analysis how can I be so confident in a publication to say that is the TSS, because there are lots of programs, but I get different results with each one of them!!

something else;
in promoter predictions, I used mat inspector and it predicted a 617bp sequence as the promoter of my gene, and part of that involves the first exon (84bp long) and 20bp from the following intron, is that possible?
And in most papers they start experiments with a 2 - 2.5kbp fragment upstream the TSS or ATG.......but I don't know why if you search for TF binding sites you find them in that sequence and even upstream from that (at least that's what I found with ALIBABA) By the way, TBP binds to the TATA box?
I also read a paper in which they found the highest activity at the end of the first intron next to exon 2...I don't guess how transcription is performed in that case!!!!

looking forward for your answer

b

-biotech!-

Hi biotech,
Any program that finds TSS solely based on sequence features is deemed not to give accurate results. TSS prediction has to rely on experimental evidence such as on EST, cloned transcripts. For many genes there is not a single TSS. You can determine TSS of a gene in many genome databases by checking the far 5' end of EST or mRNA alignment. Also the UCSC genome browser provides the FirstEF track which is also useful.

Best of all, the DBTSS (database of TSS) provides TSS prediction based on experimentally determined 5'-end clones.

-pcrman-