Question

DESEq2 Error: rows did not converge

0

Entering edit mode

andrebolerbarros ▴ 20

@andrebolerbarros-16788

Last seen 6 months ago

Portugal

Hi everyone,

I am running DESeq2 on a public available dataset of RNASeq of mouse immune cells. I am getting the error:

75 rows did not converge in beta, labelled in mcols(object)$betaConv. Use larger maxit argument with nbinomWaldTest

After checking the rowSums for the lines which did not converged (via which(mcols(dds)$betaConv==F)), I realized that the specific rows do not have low counts.

What could be the cause for this? I am using the Cell Type as variable of interest and, in this, I have quite a few groups with only two samples. Could that be the cause?

Addition: I have filtered the genes by using:

filter <- (rowSums(counts(dds)) >= 5) >= 2
dds[filter,]

but got the following error:

Error in estimateSizeFactorsForMatrix(counts(object), locfunc = locfunc,  : 
  every gene contains at least one zero, cannot compute log geometric means

DESeq2 • 1.1k views

ADD COMMENT • link 17 months ago andrebolerbarros ▴ 20

score 0 · Answer 1 · 2023-06-16

0

Entering edit mode

Michael Love 43k

@mikelove

Last seen 4 days ago

United States

Can you print the colData?

ADD COMMENT • link 17 months ago Michael Love 43k

0

Entering edit mode

colData(dds)

DataFrame with 191 rows and 8 columns
                                         Organ   Cell.Type     Stimulus         Sex                Misc1
                                      <factor>    <factor>  <character> <character>          <character>
B.Fo.Sp.1                          Spleen          B Cells Unstimulated        Male   cell type: B cells
B.Fo.Sp.2                          Spleen          B Cells Unstimulated        Male   cell type: B cells
B.Fo.Sp.3                          Spleen          B Cells Unstimulated        Male   cell type: B cells
B.Fo.Sp.4                          Spleen          B Cells Unstimulated        Male   cell type: B cells
B.FrE.BM.1                         Bone Marrow     B Cells Unstimulated        Male   cell type: B cells
...                                        ...         ...          ...         ...                  ...
NKT.19.8.TCRb.CD1daGalCerTet..Lv.1      Liver  NKT Cell    Unstimulated        Male cell type: ab T Cell
NKT.19.8.TCRb.CD1daGalCerTet..Lv.2      Liver  NKT Cell    Unstimulated        Male cell type: ab T Cell
NKT.19.8.TCRb.CD1daGalCerTet..Lv.3      Liver  NKT Cell    Unstimulated        Male cell type: ab T Cell
T.4.19.8.TCRb.CD4..Sp.1                 Spleen CD4 T Cells Unstimulated        Male cell type: ab T Cell
T.4.19.8.TCRb.CD4..Sp.2                 Spleen CD4 T Cells Unstimulated        Male cell type: ab T Cell
                                                    Misc2       Age LibrarySize
                                              <character> <integer>   <numeric>
B.Fo.Sp.1                              Splenic Follicular         6     6841754
B.Fo.Sp.2                              Splenic Follicular         6     4146448
B.Fo.Sp.3                              Splenic Follicular         6     7819577
B.Fo.Sp.4                              Splenic Follicular         6     4005546
B.FrE.BM.1                         Bone Marrow Fr. E (n..         6     6485032
...                                                   ...       ...         ...
NKT.19.8.TCRb.CD1daGalCerTet..Lv.1              iNKT Cell         6     5141838
NKT.19.8.TCRb.CD1daGalCerTet..Lv.2              iNKT Cell         6     6244931
NKT.19.8.TCRb.CD1daGalCerTet..Lv.3              iNKT Cell         7     6052795
T.4.19.8.TCRb.CD4..Sp.1                        CD4 T Cell         6     4320606
T.4.19.8.TCRb.CD4..Sp.2                        CD4 T Cell         6     3955050

ADD REPLY • link 17 months ago andrebolerbarros ▴ 20

0

Entering edit mode

The fact that every gene contains a zero is a problem for the default DESeq2 routine. Is this bulk RNA-seq? It is a bit strange to have no genes with positive counts. Maybe you could subset to groups of samples that you intend to compare.

ADD REPLY • link 17 months ago Michael Love 43k

0

Entering edit mode

It's most likely due to my many different groups and cell types. I have decided to subset into specific groups (organs & cell types) before comparing.

ADD REPLY • link 17 months ago andrebolerbarros ▴ 20