A proteogenomic approach to discover novel lncRNA-derived microproteins and their potential clinical utility in hepatocellular carcinoma

Figures

• 

  FIG. 1. Identification of three lists oflncORFs using high-quality Ribo-seq datasets and de novo ORF identifiers.A, flowchart of the study design and discovery analysis. B, bar chart demonstrating the overall reads distribution over 27 to 30 nt read lengths across filtered Ribo-seq datasets. C, overall frame reads distribution of the filtered Ribo-seq datasets. D, metagene profile showing triplet periodicity of P-site distribution in annotated ORFs. E, summary of lncORFs and parent lncRNAs predicted by RiboCode, Ribo-TISH, ribotricer de novo ORF identifiers. F. Kernel Density Estimation (KDE) showing length distribution of identified lncORFs between 15 and 300 nt g. P-site distribution showing triplet periodicity in MTLN-ORF and NCBP2AS2-ORF (Wilcoxon Rank Sum Test; ∗∗∗, p < 0.001).

• 

  Fig. 2. PSM group FDR control for optimization of the integrated lncORF database.A, Venn diagram showing the three overlapping lncORF lists and potential proteogenomic lncORF databases (database 1: lncORFs detected in at least two identifiers; database 2: lncORFs detected in at least one identifier; database 3: all ORFfinder lncORFs). B, flowchart of the Group FDR estimation and the Two-pass search as statistical strategies in selecting lncORF databases (FDR threshold 1%). C, Box-and-line plots demonstrating the number of putative lncORF-sourced MS/MS spectra identified using each lncORF database: Gao et al. (2019) (left) and Jiang et al. (2019) (right). D, bar charts showing the number of unique peptides, potential SAP peptides, and potential miss-cleaved peptides detected in Gao et al. (2019) (left) and Jiang et al. (2019) (right) datasets.

• 

  Fig. 3. RPF tracks and matched mass spectra of three lncPep examples. RPF tracks (left) and representative unique peptide MS spectra (right) of PPIAP79 lncORF1 (top), POTEKP lncORF2 (middle), and HNRNPA1P36 lncORF1 (bottom).

• 

  Fig. 4. 104lncPeps detected in HCC patient proteomic datasets.A, Sankey diagram visualizing traffic flow of the identified lncPeps between publication origin of datasets, databases, ORF identifiers, and biotypes annotated in GENECODE and NONCODE projects. B, box plots displaying the “Importance” Z scores of physicochemical properties of translated lncORFs as predictors of MS detectability, using Boruta’s algorithm (higher “Importance” Z-score indicates stronger predictor). C, box plots showing in silico tryptic peptides number (left), GEOR030105 score (middle), and lysine ratio (right) in MS peptides-matched (match) and MS peptides-unmatched (non-match) lncORF groups (Wilcoxon Rank Sum Test). D, predicted Pfam domains and subcellular localizations of the lncPeps. E, detection rates of lncPeps across cohort samples.

• 

  Fig. 5. Improved performance of the HCC recurrence prediction model by combining lncPep expression information. AUC curves of models based on canonical protein expression, lncPep expression, or Merged for (A) tumor detection, (B) survival prognosis, and (C) recurrence prognosis. Red color denotes AUC values significantly different between Merge and Protein models as detected by Likelihood Ratio Test.

• 

  Fig. 6. Newly discoveredlncPeps are associated with tumor, patient survival, and recurrenceA, heatmap shows the odds ratio (OR) for tumor, hazard ratio (HR) for overall survival prognosis, and HR for recurrence prognosis, mapped against each newly discovered lncPep (left) and sourced lncRNA (right). Statistical significance was assessed using logistic regression or Cox regression; ∗, p < 0.05. B, forest plots showing the association between expression of lncPeps (PPIAP79 lncORF1 and SEPTIN7P8 lncORF1) and OR for tumor in Jiang et al. 2019 and Gao et al. 2019. C, box plots showing expression of PPIAP79 lncORF1 and SEPTIN7P8 lncORF1 in non-tumor tissue (Non-T) and tumor tissue (T). D, box plot showing expression of HNRNPA1P36 lncORF1 in non-tumor and tumor tissues, and Kaplan–Meier (KM) analysis of overall survival probability in patients stratified by HNRNPA1P36lncORF1 expression (upper vs. lower quartile). E, box plot showing expression of POTEKP lncORF2 in Non-tumor and tumor tissues, and KM analysis of recurrence prognosis. Statistical significance was assessed using the Log-rank test. DFI, Disease-Free Interval; DSS, Disease-Specific Survival; OS, Overall Survival; PFI,…