PEPPRO: quality control and processing of nascent RNA profiling data

Abstract

Nascent RNA profiling is growing in popularity; however, there is no standard analysis pipeline to uniformly process the data and assess quality. Here, we introduce PEPPRO, a comprehensive, scalable workflow for GRO-seq, PRO-seq, and ChRO-seq data. PEPPRO produces uniformly processed output files for downstream analysis and assesses adapter abundance, RNA integrity, library complexity, nascent RNA purity, and run-on efficiency. PEPPRO is restartable and fault-tolerant, records copious logs, and provides a web-based project report. PEPPRO can be run locally or using a cluster, providing a portable first step for genomic nascent RNA analysis.

Fig. 1

PEPPRO steps for genomic run-on data. PEPPRO starts from raw sequencing reads and produces a variety of quality control plots and processed output files for more detailed downstream analysis

Fig. 2

PEPPRO test set data table and signal tracks. a Table showing the attributes of samples collected for our test set. Complete metadata is available from the PEPPRO website. b Read count normalized signal tracks from published data are visualized within a browser (Scale is per 1M)

Fig. 3

RNA integrity is assessed with degradation ratios and insert sizes. a Schematic illustrating intact versus degraded libraries. b Degradation ratio for test samples (HelaS3 GRO sample could not be calculated; Values less than dashed line (1.0) are considered high quality). c–f Insert size distributions for: c, a degraded single-end library; d, a degraded paired-end library; e, a non-degraded single-end library; and f, a non-degraded paired-end library (orange shading represents highly degraded reads; yellow shading represents partially degraded reads)

Fig. 4

Library complexity is measured with unique read frequency distributions and projections. a Schematic demonstrating PCR duplication and library complexity (dashed line represents completely unique library). b Library complexity traces plot the read count versus externally calculated deduplicated read counts. Deduplication is a prerequisite, so these plots may only be produced for samples with UMIs. Inset zooms to region from 0 to double the maximum number of unique reads. c The position of curves in panel b at a sequencing depth of 10 million reads (dashed line represents minimum recommended percentage of unique reads)

Fig. 5

Nascent RNA purity is assessed with the exon-intron ratio. a Schematic demonstrating mRNA contamination calculation. redX represents the exclusion of the first exon in the calculation. b Median mRNA contamination metric for test set samples (shaded region represents recommended range (1-1.8)). c Histogram showing the distribution of mRNA contamination score across genes in the K562 PRO-seq sample. d As in panel c for a GRO-seq library. e mRNA contamination distribution for K562 PRO-seq spiked with 30% K562 RNA-seq. f mRNA contamination distribution for HelaS3 GRO-seq is comparable to the 30% RNA-seq spike-in sample

Fig. 6

Run-on efficiency is measured with pause indices. a Schematic demonstrating pause index calculation. b Pause index values for Drosophila melanogaster GRO-seq libraries with (GSM577247) or without sarkosyl (GSM577248). c The histogram of pause index values is shifted to the right upon addition of sarkosyl in GRO-seq libraries. d Pause index values for test set samples (Values above the dashed line are recommended). e High pause index identified in H9 treated PRO-seq. f Low pause index from HelaS3 GRO-seq

Fig. 7

Fraction of reads in genomic features. a K562 PRO-seq represents a “good” cumulative fraction of reads in features (cFRiF) and fraction of reads in features (FRiF) plot. b K562 PRO-seq with 90% K562 RNA-seq spike-in represents a “bad” FRiF/PRiF

Fig. 8

Differential analysis with the PEPPRO counts matrix. a MA plot between H9 DMSO versus H9 200nM romidepsin treated PRO-seq libaries (dots = genes; top 10 most significant genes labeled; n=3/treatment). b Most significantly differential gene count differences. c Read count normalized signal tracks from the differential analysis (Scale is per 1M)

Fig. 9

Recommendation table. Based on our experience processing both high- and low-quality nascent RNA libraries, these are our recommended values for high-quality PRO-seq libraries