Extended Data Figure 2: Validation of PAL-seq performance.

a, Evidence against non-specific RNA degradation. Plotted are nucleotide identities at the positions immediately upstream of poly(A) tags that both mapped uniquely to the genome (or standards) and ranged from 22–30 nucleotides in length (a range chosen to be long enough to enable mostly unique mapping to the genome, yet short enough to include enough 5′ adaptor nucleotides in a 36-nucleotide read to clearly identify the 5′ end of the tag). Frequencies were normalized to the aggregate nucleotide composition of positions 23–31 in either uniquely genome- or standard-mapping tags that extended the full length of the reads (36 nucleotides). Because RNase T1 cuts after Gs, the nucleotide preceding each 22–30-nucleotide tag was expected to be G, unless the mRNA had been cut for some other reason. The high frequency of G indicated that most mRNA fragments had not been cut for other reasons, which also implied that for these samples the poly(A) tails had also remained intact. We are unable to explain the high signal for an upstream U or C in some samples. Nonetheless, the frequency of an upstream A was low, which indicated that there had been little cleavage after As, again implying that the poly(A) tails had remained intact. In the A. thaliana leaf analysis, for which the raw reads had the first base removed, estimation of RNA integrity was performed with length ranges shortened by one nucleotide (for example, informative poly(A) tags were 21–29 nucleotides long). b, Consistent results from similar samples or biological replicates. Plotted are the relationships between average poly(A)-tail lengths generated using HeLa total RNA or RNA from a cytoplasmically enriched lysate (left), between average poly(A)-tail lengths generated using S. cerevisiae total RNA or RNA from a cytoplasmically enriched lysate (sample 1 and 2, respectively; middle), and between average poly(A)-tail lengths generated using cytoplasmically enriched lysates from two different 3T3 cell lines (right). Although the 3T3 lines were each engineered to express a miRNA (either miR-1 or miR-155), the miRNA was not induced in the cells used for this comparison. NM_001007026 fell outside the plot for HeLa, and YDL080C fell outside the plot for yeast.