[BioC] SAM and siggenes packages

Tue Apr 27 17:04:05 CEST 2004

So I did it:

here are the last lines

[8347] 0.084854730 0.148227947 0.106908754 0.125831287 0.077950147 
0.211227137
[8353] 0.152660230 0.045091101 0.096348186 0.044063845 0.108855645 
0.092417888
[8359] 0.118671164 0.027960521 0.043253745 0.046361978 0.154575781 
0.152634900
[8365] 0.074517075 0.229593012 0.082204362 0.206090376 0.045857296 
0.094095230
[8371] 0.103852101 0.233859308 0.258523505 0.103864178 0.122385598 
0.167787555
[8377] 0.168447995 0.095190885 0.095376051 0.049927168 0.053458959 
0.102979050
[8383] 0.183686758 0.162928860 0.101409564 0.055026031 0.047275238 
0.299989073
[8389] 0.048762658 0.114255858 0.213818969 0.065428394 0.012298578 
0.061579019
[8395] 0.050404750 0.075257078 0.084360630 0.199929845 0.177594221 
0.157630579
[8401] 0.180217253 0.112000120 0.218196589 0.152459139 0.144681365 
0.073595428
[8407] 0.108158061 0.013548077 0.149060865 0.123811131 0.308569776 
0.217594634
[8413] 0.071042281 0.099095831 0.029147566 0.044104928 0.109420584 
0.116411330
[8419] 0.121903308 0.053624854 0.159189549 0.203203948 0.128026072 
0.142591910
[8425] 0.140158370 0.095307900 0.230283970 0.182668313 0.159248974 
0.215344233
[8431] 0.128684507 0.126580411 0.067487608 0.041372752 0.076218630 
0.118931153
[8437] 0.292739641 0.058138725 0.057913610 0.108836945 0.125646992 
0.194450518
[8443] 0.092050998 0.926009673 0.164894890 0.304727643 0.415619179 
0.174991675

So standard deviations seems not to be equal to zero....

I tried also with another set of expression data (golub) using the first 
four columns, five six and 12 and gave a similar result.

It's working (one class) until five replicates and gave a similar error 
that I got with my data if I worked with four replicates or less ?

Strange ?

Thanks

Best regards

Dr Christophe Perin

Holger Schwender a écrit:

>It seems that each gene has the same value for each of the four replicates
>(because of the error message). Otherwise all the expression values of the
>genes have to be NAs. Please take a look on the standard deviations by 
>
>  
>
>>rs.out<-rs.cal(salt_stress,1:4,B=1)
>>rs.out$s
>>    
>>
>
>Are they all zero? If yes, then all genes have the same value for each of
>their expression levels.
>
>  
>
>>Hi,
>>
>>Thanks.
>>I download and install the new version of siggenes. I also added the two 
>>lines of code to set s0 as the median of the standard deviations of the 
>>gene and it's working now.
>>But, I still have a pb: I played with 6 replicates as a test but I 
>>planned to generate only 4 replicates (2 dye swap) for  each step of my 
>>time course experiment. So if I try with only 4 replicates instead 6 I 
>>gave the output as follow:
>>
>>SAM Analysis for the one-class case.
>> 
>>Warning: There are 8448 genes which have variance Zero or no non-missing 
>>values.
>>         The d-value of these genes is set to NA.
>> 
>>There are 8448 missing d values.
>> 
>>Error in "[<-"(*tmp*, int, value = numeric(0)) :
>>        nothing to replace with
>>In addition: Warning messages:
>>1: no finite arguments to min; returning Inf
>>2: no finite arguments to max; returning -Inf
>>
>>with the following code:
>>
>>a<-median(rs.cal(salt_stress,1:4,B=1)$s)
>>sam.output<-sam(maM(salt_stress,salt_cl,s0=a)
>>
>>Again, any idea ? (I tried with only 5 replicates and it's working ???)
>>
>>
>> 
>>
>>
>>Holger Schwender a écrit:
>>
>>    
>>
>>>Hi,
>>>
>>>the warning message says that all the expression values of each of the
>>>      
>>>
>>353
>>    
>>
>>>genes are either the same or NAs.
>>>
>>>The error message usually occurs in the computation of the fudge factor
>>>      
>>>
>>when
>>    
>>
>>>there are less than 101 different d values. This "bug" will be fixed in
>>>      
>>>
>>the
>>    
>>
>>>next version of siggenes. You can avoid this error message by setting s0
>>>      
>>>
>>in
>>    
>>
>>>sam(...) to a reasonable value, or by doing something like
>>>
>>> 
>>>
>>>      
>>>
>>>>a<-median(rs.cal(salt_stress,1:6,B=1)$s) 
>>>>sam(salt_stress,salt_cl,s0=a)
>>>>   
>>>>
>>>>        
>>>>
>>>This will specify the fudge factor as the median of the standard
>>>      
>>>
>>deviations
>>    
>>
>>>of the genes.
>>>
>>>Since you have >8000 genes this error message shouldn't actually occur.
>>>      
>>>
>>So
>>    
>>
>>>another idea is that you might have an old version of siggenes in which
>>>      
>>>
>>the
>>    
>>
>>>one-class analysis did not work correctly. If you don't have the siggenes
>>>version of the developmental section of Bioconductor, you should download
>>>this version (1.0.6).
>>>
>>>Best,
>>>Holger
>>>
>>> 
>>>
>>>      
>>>
>>>>Hi,
>>>>
>>>>I'am new on this mailing list and I probably have a 'stupid' question.
>>>>I'am doing SAM analysis for the one-class case on a 6 replicates per 
>>>>8448 gene matrix.
>>>>I get each time this message
>>>>
>>>>SAM Analysis for the one-class case.
>>>>
>>>>Warning: There are 353 genes which have variance Zero or no non-missing 
>>>>values.
>>>>        The d-value of these genes is set to NA.
>>>>
>>>>Error in var(v) : missing observations in cov/cor
>>>>
>>>>Any idea ?
>>>>
>>>>Here the code:
>>>>
>>>>sam.output<-sam(salt_stress,salt.cl)
>>>>
>>>>where salt.cl<-c(rep(1,6))
>>>>and salt_stress a matrix (6 col per 8448 row)
>>>>
>>>>
>>>>thanks
>>>>
>>>>
>>>>
>>>>
>>>>Thanks
>>>>
>>>>_______________________________________________
>>>>Bioconductor mailing list
>>>>Bioconductor at stat.math.ethz.ch
>>>>https://www.stat.math.ethz.ch/mailman/listinfo/bioconductor
>>>>
>>>>   
>>>>
>>>>        
>>>>
>>> 
>>>
>>>      
>>>
>>    
>>
>
>  
>

	[[alternative HTML version deleted]]