Dexseq warnings: Failed to fit dispersion for gene

[capricy]

Hi, there,

I managed to construct the exon count subject. When I tried to use the function: estimateDisperions, I got a lot of warnings. Could anybody explain those to me?

Thanks a lot

----------

> warnings()
Warning messages:
1: In .local(object, ...) :
Exons with less than 10 counts will not be tested. For more details please see the manual page of 'estimateDispersions', parameter 'minCount'
2: In .local(object, ...) :
Genes with more than 70 testable exons will be omitted from the analysis. For more details please see the manual page of 'estimateDispersions', parameter 'maxExon'.
3: In .local(object, ...) : Failed to fit dispersion for gene CUFF.10004
4: In dispBeforeSharing[rows] <- disps :
number of items to replace is not a multiple of replacement length
5: In .local(object, ...) : Failed to fit dispersion for gene CUFF.10007
6: In dispBeforeSharing[rows] <- disps :
number of items to replace is not a multiple of replacement length
7: In .local(object, ...) : Failed to fit dispersion for gene CUFF.10010
8: In dispBeforeSharing[rows] <- disps :
number of items to replace is not a multiple of replacement length
9: In .local(object, ...) : Failed to fit dispersion for gene CUFF.10014
10: In dispBeforeSharing[rows] <- disps :
number of items to replace is not a multiple of replacement length


my inputs:
"CUFF.10002:001" 0 0 0 0 0 0 0 0
"CUFF.10002:002" 0 0 0 0 9 10 3 0
"CUFF.10004:001" 0 0 0 0 0 0 0 0
"CUFF.10004:002" 2 1 0 0 2 0 2 0
"CUFF.10004:003" 1 0 0 0 1 0 0 0
"CUFF.10004:004" 3 2 0 0 2 0 4 0
"CUFF.10004:005" 0 0 0 0 1 0 3 0
"CUFF.10004:006" 2 2 2 0 7 0 4 1
"CUFF.10004:007" 6 12 35 0 46 6 15 6
"CUFF.10004:008" 5 6 48 0 24 15 7 1
"CUFF.10007:001" 0 0 0 0 0 0 0 0
"CUFF.10007:002" 0 0 0 0 0 0 0 0
"CUFF.10007:003" 2 0 0 1 0 0 1 2
"CUFF.10007:004" 14 4 11 25 19 30 25 24
"CUFF.10007:005" 27 7 64 29 37 5 84 8
"CUFF.10007:006" 3 0 0 0 0 1 2 2
"CUFF.10007:007" 1 3 3 2 1 43 11 2
"CUFF.10007:008" 68 57 84 48 93 138 103 45
"CUFF.10007:009" 71 74 94 53 104 101 80 52
"CUFF.10007:010" 91 102 195 72 125 136 118 77
"CUFF.10007:011" 37 21 67 28 40 11 39 22
"CUFF.10007:012" 26 5 44 29 23 15 33 14
"CUFF.10010:001" 0 0 0 0 0 12 0 0
"CUFF.10010:002" 143 62 37 54 153 17 188 84
"CUFF.10010:003" 216 142 67 35 155 117 204 184
"CUFF.10010:004" 266 177 179 54 281 81 286 290
"CUFF.10010:005" 240 144 255 47 251 39 253 282
"CUFF.10010:006" 239 133 367 23 190 30 303 293
"CUFF.10010:007" 325 273 327 18 291 30 504 405
"CUFF.10010:008" 236 261 120 33 290 4 423 336
"CUFF.10010:009" 186 213 87 36 235 5 351 286
"CUFF.10010:010" 151 160 68 37 182 4 274 222
"CUFF.10010:011" 252 226 111 72 288 7 409 322
"CUFF.10010:012" 130 127 90 45 165 5 237 157
"CUFF.10010:013" 265 293 208 138 361 5 473 311
"CUFF.10010:014" 197 242 186 143 312 0 382 235
"CUFF.10010:015" 91 100 85 68 140 0 178 113
"CUFF.10010:016" 141 165 118 92 197 2 262 186
"CUFF.10010:017" 38 49 22 17 44 1 54 58
"CUFF.10010:018" 16 17 8 0 15 1 12 25
"CUFF.10010:019" 0 6 0 0 0 0 0 3
"CUFF.10010:020" 0 6 0 0 1 0 0 1
"CUFF.10010:021" 0 2 2 0 1 0 1 2
"CUFF.10010:022" 0 1 0 0 1 0 1 2
"CUFF.10010:023" 0 0 0 0 1 0 1 1
"CUFF.10011:001" 29 23 64 119 69 299 52 22
"CUFF.10014:001" 3 1 0 0 0 0 16 0
"CUFF.10014:002" 210 105 188 0 113 69 402 305
"CUFF.10014:003" 192 74 143 1 122 64 341 263
"CUFF.10014:004" 354 186 232 15 285 84 623 493
"CUFF.10014:005" 641 476 408 44 631 213 1127 972
"CUFF.10014:006" 302 256 230 27 366 145 568 519
"CUFF.10014:007" 210 220 223 17 306 149 503 430
"CUFF.10014:008" 90 95 122 1 147 80 241 211
"CUFF.10014:009" 105 72 120 13 134 85 239 205
"CUFF.10014:010" 47 22 39 10 34 30 73 78
"CUFF.10014:011" 19 12 12 1 8 5 17 20
"CUFF.10016:001" 3 0 12 37 18 65 8 4
"CUFF.10016:002" 0 0 0 0 0 0 1 0

[dpryan]

What are the size factors for the libraries and what's the design? I would guess that one or more of the libraries are much smaller than the others. Having said that, a number of those do seem to have >100x difference in counts between the sample with the fewest counts and the median.

[capricy]

Here are size factors:

> sizeFactors(ecs.m)
../ATCACG/genome_mapping/tophat_out/accepted_hits.exonCounts
1.5774367
../GATCAG/genome_mapping/tophat_out/accepted_hits.exonCounts
1.0939807
../CGATGT/genome_mapping/tophat_out/accepted_hits.exonCounts
1.4076148
../TAGCTT/genome_mapping/tophat_out/accepted_hits.exonCounts
0.3917576
../ACAGTG/genome_mapping/tophat_out/accepted_hits.exonCounts
1.7156390
../TTAGGC/genome_mapping/tophat_out/accepted_hits.exonCounts
0.8036881
--------------

do you think 0.39~1.71, is huge difference in terms of library size?

[areyes]

Hi capricy,

One question, did you updated your version of R and DEXSeq, as we discussed in the mailing list? I think this would be very helpful for your analysis.

Alejandro

[capricy]

I am still using R2.15 (I need to figure out who is in charge...). But I figured out where those warning messages came from: after subsetting the data, I would need to do droplevels, which I didn't. So all these warnings are gone.

My next question is about the plot/visualization: since I am doing 3 conditions comparison, MA, MM, and MP, and the results showed the log2fold(MM/MA) and log2fold(MP/MA), but the plot seems to only display the comparison between MP/MA. I wonder how I would be able to show comparison between MM/MA?

Thanks a lot for your help!