RPKM Expression

Analyze RNA-Seq expression in vicinity of integration events.

Evaluate the expression of case versus controls, of which there are typically many.

• In L_Expression stage, obtain expression data from RNA-Seq BAMs directly (RPKM format).
• An alternate workflow is illustrated in stage M_RSEM_Expression which uses pre-processed RSEM expression data, with no need for sequence (RNA-Seq) data.

Either L_Expression or M_RSEM_Expression stage is required for BPS expression plots, but not necessary for structure plots.

1. For each case, create BED file which indicates regions of interest where RPKM data will be evaluated. Regions of interest are all exons belonging to genes within some distance - here, 1Mbp - of integration event.
2. Analyze RNA-Seq data for both cases (sample with integration event in given region) and controls (all other samples with same cancer type as case) to obtain RPKM, a normalized count of RNA-Seq reads mapped to regions of interest. Note that this step reads RNA-Seq BED files.
   • Raw RNA-Seq reads are counted for each exon using bedtools coverage. Per-exon RPKM is evaluated as ($10^{9}$ * R)/(N * L), where R is the number of raw reads mapped to an exon, N is the count of total mapped reads, and L is exon length.
3. Evaluate expression of case exons vs. control. Combine exon expression to obtain per-gene p-value which quantifies degree to which gene expression differs in case vs. controls.
   • Specifically, we report the probability that expression of case is drawn from the same distribution as expression of controls.
   • See $BPS_CORE/src/analysis/ExonExpressionAnalyzer.R
   • Algorithm details available here.

The third step can work with other expression data, e.g. RSEM data as illustrated in M_RSEM_Expression.