Integrating transcriptomic and proteomic data for accurate assembly and annotation of genomes

Abstract

Complementing genome sequence with deep transcriptome and proteome data could enable more accurate assembly and annotation of newly sequenced genomes. Here, we provide a proof-of-concept of an integrated approach for analysis of the genome and proteome of Anopheles stephensi, which is one of the most important vectors of the malaria parasite. To achieve broad coverage of genes, we carried out transcriptome sequencing and deep proteome profiling of multiple anatomically distinct sites. Based on transcriptomic data alone, we identified and corrected 535 events of incomplete genome assembly involving 1196 scaffolds and 868 protein-coding gene models. This proteogenomic approach enabled us to add 365 genes that were missed during genome annotation and identify 917 gene correction events through discovery of 151 novel exons, 297 protein extensions, 231 exon extensions, 192 novel protein start sites, 19 novel translational frames, 28 events of joining of exons, and 76 events of joining of adjacent genes as a single gene. Incorporation of proteomic evidence allowed us to change the designation of more than 87 predicted "noncoding RNAs" to conventional mRNAs coded by protein-coding genes. Importantly, extension of the newly corrected genome assemblies and gene models to 15 other newly assembled Anopheline genomes led to the discovery of a large number of apparent discrepancies in assembly and annotation of these genomes. Our data provide a framework for how future genome sequencing efforts should incorporate transcriptomic and proteomic analysis in combination with simultaneous manual curation to achieve near complete assembly and accurate annotation of genomes.

Figure 1.

Overview of pipeline used for correction of genome annotation and genome assembly using transcriptomic and proteomic data. MS/MS spectra, which did not assign to the known protein database, were further searched against six-frame translated genome, three-frame translated transcripts, and Anopheles gambiae protein database. Further analysis of these peptides resulted in identification of novel protein-coding genes and revised gene annotations, which were compared against 15 other Anopheline species.

Figure 2.

Schematic representation of the workflow and summary of proteomic data. (A) Adult tissues and developmental stages of the Indian strain of An. stephensi that were dissected and processed for transcriptomic or proteomic analysis. (B) Revised annotation of An. stephensi genome based on RNA-seq evidence. The numbers represent the junctional reads identified in each tissue, and the two transcript models shown are splice variants identified based on RNA-seq data. (C) Broad overview of mass spectrometry–based proteomic analysis of multiple tissues. (D) Median mass error of the peptide spectral matches identified in the study. (E) Total number of peptides identified against AsteI2 and AsteS1 assembly. (F) Total number of genome search-specific peptides identified against the two assemblies. (G) Insertion of five small scaffolds in genome gap regions of scaffold00004.

Figure 3.

Reannotation of the An. stephensi genome based on transcriptomic and proteomic evidence. (A) Alignment of RNA-seq data against the 15.4-kb-long mitochondrial genome reveals transcript evidence (colored arrows) for 13 mitochondrial genes, of which seven were also identified at the protein level (red lines). (B) Insertion of a nucleotide base in mitochondrial genome based on RNA-seq evidence. (C) Identification of a novel gene in the AsteI2 and AsteS1 assemblies. (D) Alternate protein start site based on the presence of an upstream N-terminally acetylated peptide.

Figure 4.

Comparison of annotations across 16 Anopheline genomes. (A) A schematic representation comparing 100 novel representative annotations identified in An. stephensi with 15 other Anopheline species. The 100 novel events identified in An. stephensi were selected at random and mapped across the 15 Anopheline genomes to check for orthologous sequences in these species. (B) Comparison of one such gene (annotation shown in blue in An. stephensi) provided evidence for revised genome annotation in An. quadriannulatus, An. Arabiensis, and An. albimanus.

Figure 5.

Translational evidence for predicted noncoding RNAs. (A) Multiple peptides from various tissues were identified that corresponded to a computationally predicted noncoding RNA. (B) Relative expression of proteins encoded by computationally predicted noncoding RNAs that were detected across tissues.