run_drimseq stops with some datasets

[MRonzio]

Describe the bug Hi, thanks for your contribution. It is a very useful and documented package. Unfortunately, running the DTU analysis on some datasets it gives error: Error in constrOptim(theta = prop_init[-q], f = dm_likG, grad = dm_scoreG, : initial value is not in the interior of the feasible region. In some cases, removing some samples and/or some transcripts solves the issue but I couldn't understand exactly the rationale. For my understanding, constrOptim function would work adding a value to the theta parameter, but I am not perfectly sure where to add it. Any hint? Thanks To Reproduce

dturtle <- run_drimseq(counts = cts_sub, tx2gene = tx2gene, pd=pd, id_col = "sample_file",
                    cond_col = "cond", cond_levels = c("cond1", "cond2"),
                    filtering_strategy = "bulk", BPPARAM = biocpar)

Adding pseudovalue=T still gives error.

Please complete the following information:

• OS: Ubuntu 22.04 LTS
• R-Version 4.2 and 4.3
• DTUrtle Version 1.0.2

[MRonzio]

For anyone getting the same issue, I was able to solve by adding a pseudovalue (0.01) to the cts_sub matrix.