Published today (NAR GAB). Stan used for outlier detection for genetic analyses (RNA sequencing)

Publicity
1

With great pleasure that I announce that our article " Probabilistic outlier identification for RNA sequencing generalized linear models" has been published in NAR Genomics and Bioinformatics.

  • This article showcases our iterative strategy for outlier detection
  • We show the utility of Variational Bayes, with no significant loss in accuracy compared to HMC (for our model)

Thanks a lot to all the Stan community, an in particular to co-authors that made this possible (including @martinmodrak and @avehtari), and @Bob_Carpenter for the precious feedback.

7 Likes
2

This looks very interesting. Do you think a similar approach would work for detecting F_{ST} outliers in SNP data? It sounds a lot better than the approach we tried (JASA 2009: https://www.tandfonline.com/doi/abs/10.1198/jasa.2009.0010)

3

Dear Kent,

interesting paper. This iterative approach is generally applicable for models for which (i) generated quantities can be produced, (ii) single observations can be omitted (or imputed) from the likelihood calculation, (iii a plus) a truncated distribution can be modeled.

With a quick read I believe it is possible to adapt this approach to your problem.