Solid Mapping

[chouf]

hello,

I would like to know, When you have reads about 35/50 Bp, And you want to see if the gene is expressed or not.

How many reads must have it to show that the gene is expressed?

Thanks

Chouf

[Richard Finney]

One high "mapability" read (i.e. "it really does go to that location") is evidence that that gene is expressed.

[chouf]

So for example on this pictures, you have a gene expressed or not?

May be there is an average of reads?

thanks

[mbblack]

So for example on this pictures, you have a gene expressed or not?

May be there is an average of reads?

thanks

I'm not sure if you are thinking of two different things. If you mean is a gene detected, that is was it expressed in the sampled mRNA pool, then a single unique mapped read would answer yes. As long as your mapping stringency was high (a reasonable read QV cutoff and a mapping score QV cutoff).

Would you want to base actual estimated expression levels off such genes, as in a differential gene expression analyssis? No, I would say not. In the literature, you will find people using some minimum mapped read count for inclusion in their esitmations of expression levels. So, only genes having a minimum number of uniquely mapped reads of say, greater than 5 or greater than 10 are common numbers (or so is my take from published analyses).

So there are two questions.

1. Was a gene/transcript/feature detected at all - basically presence or absence of a mappable feature?
2. Which genes have sufficient reads to provide reliable estimates of relative expression?

For 1. again if you set up your mapping well (read QV stringency and mapping QV stringency), then even a single unique mapped read is evidence for the detection of a gene or transcript.

For 2. I would not be comfortable actually using any genes with fewer than 10 mapped reads in any relative expression estimates (since this is now a question of having enough data to confidently estimate an actual level of expression, not merely the presence or absence of a feature).