Bioinformatics Seminars

Bioinformatics Seminar

Time: 10:45am Tuesdays.
Level 7 Seminar Room 2, WEHI1

10 March 2020

Topconfects ranks differentially expressed genes using confidence bounds on the fold change, with multiple testing adjustment

Paul Harrison

Bioconductor package topconfects ranks differentially expressed genes by confidence bounds on the log fold change. Highly ranked genes are those with confidently large fold change, with no need for the user to adjust parameters or apply further ad-hoc filtering. In contrast to ranking results by p-value (a common default in software), as standard errors get smaller they matter less and less to the ranking, so that with sufficient replicates the ranking is entirely by fold change.

Topconfects adjusts confidence bounds so that the 'False Coverage-statement Rate' is maintained if only the head of the ranking is selected. This adjustment shares features with FDR adjustment of p-values. Topconfects can be viewed as an alternative presentation of TREAT test results.

The method can be applied to other data types and effect sizes. If testing gives high 'significance' to uninteresting results, topconfects with a suitable choice of effect size may produce a more relevant ranking of results. Gene set enrichment will be used as an example.

Search past seminars