Warning

[maxozo]

Warning

[francois-a]

Hi,
No, this isn't a problem if it only happens rarely.

[maxozo]

Thanks for the response. What if it happens quite often?

    processing phenotype 42888/53437    * WARNING: excluding 2259 monomorphic variants
WARNING: scipy.optimize.newton failed to converge (running scipy.optimize.minimize)

    processing phenotype 42889/53437    * WARNING: excluding 2234 monomorphic variants

    processing phenotype 42890/53437    * WARNING: excluding 2240 monomorphic variants

    processing phenotype 42891/53437    * WARNING: excluding 2241 monomorphic variants

    processing phenotype 42892/53437    * WARNING: excluding 2244 monomorphic variants

    processing phenotype 42893/53437    * WARNING: excluding 2268 monomorphic variants
WARNING: scipy.optimize.newton failed to converge (running scipy.optimize.minimize)

[francois-a]

The input VCF should only include common variants. The large number of monomorphic variants suggests that your input VCF was not filtered, and this is definitely not ideal (large numbers of rare variants will potentially lead to regression outlier artifacts).

[maxozo]

thanks francois-a. Really helpful.
What filters would you suggest to be applied to vcfs before running tensorQTL?

[francois-a]

In general sample sizes are not sufficient to map variants with MAF < 0.01. So applying a MAF filter and removing monomorphic variants should be part of initial QC. The MAF filter can be applied in-sample (with the maf_threshold option), but this won't remove monomorphic variants that are all, e.g., 0/1.

[baishengjun]

I have filter the variants with MAF < 0.05, but I encounter too much the same warning above.
My sample size is only 38, does it cause this warning? and any solution to sovle this problem?

[Jesson-mark]

@francois-a
Hi,
I currently encounter the same warning but with different information when running tensorQTL with cis.map_cis function.

The logfile are:

/path/to/tensorqtl/core.py:315: RuntimeWarning: invalid value encountered in sqrt
  return 2*stats.t.cdf(-np.abs(np.sqrt(tstat2)), dof)
WARNING: scipy.optimize.newton failed to converge (running scipy.optimize.minimize)

I already filtered my genotype file by MAF>0.01 so I didn't received warning about removing monomorphic variants.
I'm running cis.map_cis to 15000 genes and this WARNING above occured 1286 times in my latest run.

I'm wondering if this warning has some effect on the beta-approximated p-value of the 15000 genes and if it may reduce the power of discoverying significant eGenes?

Any input will be highly appreciated!

JieWang