Detecting A Shift In A Vector

Question

[Patrick]

Hello guys,

I have a vector representing the enrichment in transcription factor X for a given group of genes for a given process, for 23 yeast species.

I would like to detect (regardless of the phylogenetic "order" of the species) where during evolution a shift happened in the enrichment in that TF X

Enrichment was calculated using PWM for each specie.

What I would like to do is to adress a statistically significance of that shift and then try to map the shift into a phylogenetic specie tree to localize what internal branch is concerned by this shift.

So my question is more likely an asking for help in the statics part.

Can anyone suggest a good way for detecting this shift including testing the normality of the distribution, transforming the data, calculating coreelation matrix ?? any idea is welcome

Thanks

Pat

[Ido Tamir]

So its possible that e.g. 1/2 or more or less of your yeast species show a shift. Or you could have a gradual increase across the 23 species, including a loss at some. Do you see anything interesting when you do a boxplot or stripchart? I would also ask at stats.stackexchange.com.

Answer 1

[Audriusa]

As I understand, you need a method to detect that the translation frame has shifted from its original reading frame. I think something can be done with statistical analysis of the codon usage frequency; a shifted frame may use codons that are valid in general but not very common in that organism (the codon usage statistics can be extracted from various other coding sequences of that organism, if available). Normally your shifted region will be very short as there is also a threat to hit a stop codon as well, eliminating further (and most likely any, at RNA level) translation of your protein. It is usually trivial to spot the right translation frame just because a protein sequence in this frame is not interrupted by multiple stop codons, as it does in other frames.