How to assign values to size factor (DESeq)

[kentnf]

I updated my DESeq recently, but got problem when assigning values (No. of total mapped read) to the size factor.

here is my code:

libsizes<-c( wt_rep1=18303917, wt_rep2=18947532, mu_rep1=20295971, mu_rep2=18256798)
sizeFactors(cds)<-libsizes
cds <- estimateDispersions( cds )

The R has warnings like this:

1: In log(ifelse(y == 0, 1, y/mu)) : NaNs produced
2: step size truncated due to divergence

Does anyone encounter the same problem or know the reason.

Thanks a lot!

[kentnf]

By the way, I got another warning when analysis data without replicate using DESeq.

when running below script follows the manual:
cds1<-estimateDispersions( cds1, method="blind", sharingMode="fit-only" )

I got the warning like this:
Warning message:
In parametricDispersionFit(means, disps) : Dispersion fit did not converge.

[labunit]

I am also experiencing difficulties when trying to use sizeFactors different from those estimated by estimateSizeFactors(). The local fit fails for some datasets when using custom sizeFactors but is successful when using estimateSizeFactors().
Maybe Simon could shed some light on the dependence of the local fit on the sizeFactors?

[Simon Anders]

It is rarely a good idea anyway to assign size factors manually. Especially the total number of reads is hardly ever useful. Furthermore, size factors are typically values around1 (they are relative scaling factors), so if you use your own numbers (for whatever reason), better divide them by their geometric mean before assigning them).

I should remove the explanation how to assign values to sizeFactors from the vignette to make clear that this is not something one should usually do.

[frymor]

I get this same massage without having to calculate my own size factors. I have three replica and I use the default settings of the package.

I still get this message . Is it because the size factors differs so greatly from each other?

[1] "sizeFactors( cds )"
            rep1_1                 rep1_2                 rep1_3 
           0.8318137            0.8369557            0.8329977 
            rep2_1                 rep2_2                 rep2_3
           1.2068798            1.2132170            1.2147831