Computing the ARSI score

The ARSI score was determined based to the following scheme: A given gene, transcript, or genetic region (UTR, intron, CDS, etc.) P, can be described as a sequence of nucleotides S; thus, the measure is based on the tendency of substrings in S to appear in other genetic elements, i.e. in a reference set G. Hence, computing the ARSI (G,S) score of a specified sequence (S) given a reference set of genomic elements (G) is done in two steps (see Fig. 1b): 1) For each position i in the sequence S find the longest substring S _i^j that starts in that position and appears in at least one of the sequences of the reference set G. 2) Let |S| denote the length of a sequence S; the ARSI of S is the mean length of all the substrings S _i^j, i.e. ARSI = ∑|s _i^j|/|S|.

Please note that the ARSI measure is based on a reference genome of a given organism, and therefore is not expected to be affected by various sequencing errors/biases that appear in Next Generation Sequencing (NGS) experiments. Specifically, in this study the error rate is very low for the analyzed organisms (less than 1 to 1000). As these errors distribute relatively uniformly, their effect the ARSI score is negligible: for example in E. coli the Spearman correlation between the ARSI scores and the one obtains for a simulation with uniform error rate of 1:1000 is higher than 0.99 (p < 5 · 10⁻³²³) for all 100 such randomization that were performed.