PacBio Iso-Seq library preparation and sequencing

Library construction and PacBio sequencing were performed according to the official protocol described by Pacific Biosciences (Pacific Biosciences, Menlo Park, CA, USA). Briefly, total mRNAs (~ 15 µg) of each sample was reversely transcribed into cDNA using the SMARTer™ PCR cDNA Synthesis Kit (Clontech, CA, USA) that was optimized for preparing high-quality, full-length cDNAs. The amplified cDNA products were purified for library construction using the SMRTbell template prep kit 1.0. Libraries by annealing a sequencing primer and adding polymerase to the primer-annealed template. The polymerase-bound template was bound to MagBeads and SMRT sequencing was then performed on the Pacific Bioscience Sequel System using the Sequel Sequencing kit 2.1. Sequence movie files from the two datasets were processed and analyzed through the Iso-Seq pipeline using PacBio SMRT Analysis Server v2.3.0 (http://www.pacb.com/products-andservices/ analytical-software/smrt-analysis/) to filter out polymerase reads < 50 bp and quality < 0.8 with 0 minimum full passes (Table 1). This allows for the highest yield of reads of insert (ROIs) consensus sequences going into the subsequent steps, while creating higher accuracy consensus sequences where possible. The filtered ROIs were classified into four categories: full-length non-chimeric (FLNC), full-length chimeric, non-full-length, and short reads. This is done by identifying the 50 and 30 adapters used in library preparation as well as the poly (A) tail. Only reads that contain all three in the expected arrangement and do not contain any additional copies of the adapter sequence within the DNA fragment are classified as full-length non-chimeric. Furthermore, FLNC ROIs sequences were passed through the isoform-level clustering algorithm ICE (Iterative Clustering for Error Correction). ROI sequences were used to correct errors (polish) in the isoform sequences using the Quiver software module. The polishing process of Quiver generated high-quality (HQ) isoform sequences to an expected accuracy of ≥ 99 %. Finally, the high-quality consensus transcripts of libraries from the two samples were merged and redundancy was removed using CD-HIT-EST (c = 0.95) to obtain final FL isoforms for further analysis [62].