TargetScan8 uses 3'-UTR profiles, which represent the prevalence of tandem 3' UTR isoforms of differing lengths, all of which share the same stop codon.
This representation was introduced by Nam et. al., 2014.
The 3'-UTR profiles were constructed using 3P-seq tags (with the number of tags for the 3' UTR labeled on the y-axis), which indicate the location and usage of mRNA cleavage and polyadenylation sites (Jan et al., 2011). 3P-seq tags from multiple cell lines or tissues were normalized to each other (to account for variable sequencing depth), and then aggregated into one consensus set of counts. Normalized 3P-seq tags were assigned to the representative 3' UTR of each stop codon and summed (as indicated at the left side of the profile) to provide quantification for the usage of that stop codon. This sum (2312 in the UTR profile shown above) also includes 5 pseudocounts added at at the distal end of the Gencode annotation.
Annotation of each representative 3' UTR started with the longest Gencode 3' UTR, which was sometimes extended using information from 3P-seq or other annotation sources. The 3'-UTR profile (red line) drops with each set of clustered 3P-seq tags, indicating the fraction of transcripts that includes that segment of the 3' UTR. Each TargetScan 3'-UTR profile also shows the location of the most distal end of a Gencode annotation (blue vertical line with Ensembl transcript ID).
More information on the methods used to build 3'-UTR profiles is available in Agarwal et al., 2015.