Iso-Relevance Functions - A Systematic Approach to Ranking Genomic Features by Differential Effect Size

Abstract

Abstract. It is common to measure a large number of features in parallel to identify those differing between two experimental conditions -e.g. the search for differentially expressed genes using microarrays or RNA-Seq. Ranking features by p-value allows for control of the TYPE I error, but p-values are not reliable when there are very few replicates; and investigators typically require features be ranked by "fold change"x/ȳ in conjunction with p-values. At first glance the fold change appears to be a natural quantity on which to compare the differential behavior of features. But it is highly sensitive to small values in the denominator and is problematic in how it equates changes in both small and large numbers such as a change from 1 to 2 versus a change from 100 to 200. The strategy of adjusting all values by adding one is a widely used heuristic approach to try to mitigate the problems with fold-change. However, that can be far from optimal. A systematic strategy to determine an optimal value (pseudocount) to adjust by is employed using both real and simulated benchmark data. In RNA-Seq a value of 20 appears to be close to optimal in all cases. Another strategy is to sort by differenceȳ −x, but this is problematic for comparing measurements across a wide spectum, as large differences of small values rank below proportionally smaller difference in large values. An abstract mathematical framework is introduced to describe the problem of ranking by differential effect size, enabling us to study the ranking problem in general as opposed to specific contexts such as fold-change or difference. From this framework we discovered a remarkable property of pseudocounts, in that they strike a balance between sorting by fold-change and sorting by difference. Lastly, another fundamentally different type of application is presented, which is to rank di-codons by their differential abundance in the ORFeome of different species.

Comparative evaluation of full-length isoform quantification from RNA-Seq

“…As a measure of the accuracy of each method, we compute the absolute value of the adjusted log2 fold-change ( absolute adjusted log2FC ; pseudocount of 20) 27 for estimated counts relative to the known simulated true counts. For example, if x is the true count and y is the estimated count for a particular method, we calculate the quantity of for each | | x+20 y+20 | | transcript.…”

“…To better reveal the differences between the methods we plot the percentiles of the cumulative distribution of absolute adjusted log 2 fold change (adjusted log2FC) 27 of estimated counts relative to true counts, for all expressed transcripts in all samples of each tissue (Fig. 2c, Sup.…”

“…These two distributions can be plotted together and interpreted visually -and can also be compared using the Kolmogorov-Smirnov test. The Teqila tool was used for this analysis (28). Surprisingly, the number of sibling isoforms (other isoforms of the same gene) is far less relevant than the length or number of exons, except for NRP where the number of sibling isoforms is strongly associated with accuracy.…”

Distinct vascular genomic response of proton and gamma radiation

“…The antisense signal showed no significant findings at this level; thus, only sense signal results are reported. Differences between increasing doses (q-value  0.1) were visualized by plotting the empirical cumulative distribution (eCDF) of the gene expression ratio (expression value at each dose in cGy divided by expression value at 0 cGy) as a non-parametric estimator of the underlying CDF (35).…”

Distinct vascular genomic response of proton and gamma radiation—A pilot investigation

“…The antisense signal showed no significant findings at this level; thus, only sense signal results are reported. Differences between increasing doses ( q -value ≤ 0.1) were visualized by plotting the empirical cumulative distribution (eCDF) of the gene expression ratio (expression value at each dose in cGy divided by expression value at 0 cGy) as a non-parametric estimator of the underlying CDF [38].…”