ABPEPserver: a web application for documentation and analysis of substitutants

As previously published, a reanalysis of substituants peptide expression was undertaken [13]. Briefly, Philosopher pipeline was used to detect endogenous substituants in the large-scale proteomics dataset with the same parameters as previously published [6].

Mass spectrometry-derived processed spectra (mzML) files for all independent cancer types were obtained from the CPTAC database [8]. The database of protein sequences was prepared in one of two ways. First, the human proteome with all instances of tryptophan amino acids in the proteome changed to all other amino acids except Arginine and Lysine was used as a database in the scan – referred to as database 1(fully substitutant). Second, to optimize true positives, we generated a second database (optimized database) which includes the canonical human proteome (UniPROT) with the substitutant tryptic peptides (length > 5 & < 50 amino acids) spanning tryptophan residue and tryptophan substituted to all other amino acids. The analysis of both these databases is presented separately on the server. Additional details on the FASTA file are available on the GITHUB page and the description section on the server. Briefly, MSFragger searches the mzML spectral files against the custom database for peptide detection with the following parameters; Precursor mass lower: − 20 ppm, Precursor mass upper: 20 ppm, precursor mass tolerance: 20 ppm, calibrate mass: True, Deisotoping: True, mass offset: False, isotope error: Standard, digestion: Strictly tryptic (Max. missed cleavage: 2), Variable modifications: 15.99490 M 3, 42.01060 [^ 1, 144.1021 n^ 1, 144.1021 S 1, Min Length: 7, Max Length: 50, digest mass range: 500:5000 Daltons, Max Charge: 2, remove precursor range: − 1.5, 1.5, topN peaks: 300, minimum peaks: 15, precursor range: 1:6, add Cysteine: 57.021464, add Lysine: 229.162932, among other basic parameters (Supplementary Table 1). Next, PeptideProphet validates detected peptides with the following parameters; accmass: TRUE, decoyprobs: TRUE, expectScore: TRUE, Glycosylation: FALSE, ICAT: FALSE, masswidth: 5, minimum probability after first pass of a peptide: 0.9, minimum number of NTT in a peptide: 2, among other parameters (Supplementary Table 1). Isobaric quantification was then undertaken the following parameters (bestPSM: TRUE, level: 2, minProb 0.7, ion purity cut-off: 0.5, tolerance: 20 ppm, among other parameters (Supplementary Table 1). Next, to only retain confident peptides, peptides were filtered using stringent False Discovery Rate (FDR) filtering. The following parameters were used for FDR filtering; FDR < 0.01, peptideProbability: 0.7, among other parameters (Supplementary Table 1). Next, TMT-integrator was used to create integrated reports with isobaric quantification across all samples with the following parameters (retention time normalization: False, minimum peptide probability on top of FDR filtering: 0.9, among other parameters (Supplementary Table 1).

Substitutant peptides were fetched from the reports of TMT Integrator (version 3.1.0). Using a R-script, peptides with a log2-transformed intensity score above 0 in a sample were observed as positively detected peptides in that sample. As described before [13], for intra-tumour type analysis a filter for the maximum number of samples was applied to retain peptides with higher specificity in expression, except for W > F substitutants due to their exclusive significant and specific distribution wherever significant. All tumour types have been demonstrated to be exclusive with the analysis of database 1 [13], while GBM, UCEC, and PDA did not show this exclusivity in the analysis of database 2. This optimizes the signal for gene expression correlation analysis. Furthermore, this script was used to plot bar plots depicting the cumulative number of tryptophan substitutants detected in the scans.

Eight independent human tumour-types, namely Lung Squamous Cell Carcinoma (LSCC, [14]), Clear Cell Renal Cell Carcinoma (CCRCC, [5]), Glioblastoma (GBM, [20]), Head and Neck Squamous Cell Carcinoma (HNSCC, [11]), Hepatocellular Carcinoma (HCC, [9]), Ovarian Serous Cystadenocarcinoma (OVSCC, [12]), Pancreatic Ductal Carcinoma (PDA, [3]), Uterine Corpus Endometrial Carcinoma (UCEC, [7]), were analysed to generate ABPEPserver database (Table 1) [1]. CPTAC IDS of the datasets are provided in Table1 and Supplementary Table 2.

A MySQL database was created to efficiently organize output data and avoid storage difficulties of multiple large files. With a database, data is efficiently stored and easily retrievable. For each cancer type, we stored substitutant counts. We identified individual substitutant peptides and gene cluster data from the proteomic analysis and associated gene expression. Data for tumour and adjacent normal tissue samples are made distinguishable for comparison.

ABPEPserver is a R/Shiny application which allows users to interact with and visualize our data and analysis. We implemented the R package RMySQL 0.10.23 to connect our database to the application.

Users are provided with background scientific information, methods and cancer types used in the study on the home page of the ABPEPserver. Users are provided then two options for using the web tools, viz. Analyze and tryptophan to phenylalanine (W > F) Substitutants. In the “Analyze” module, the user can explore the enrichment of substitutants and their association with molecular gene expression signatures. On the other hand, the “W > F Substitutants” module can be employed to browse individual substitutants in multiple cancer types and tumour vs adjacent normal tissue expression. Here, the information on the database used to detect the peptide is also added. The corresponding files of this module are downloadable.

ABPEPserver is a web database that serves as a platform to identify and characterize the expression of substitutants, a recently identified class of aberrant proteins with immunotherapeutic potential, across various human cancer types. In addition, ABPEPserver allows the analysis of the association of molecular gene expression signatures to the enrichment of substituants. As an example, such analysis was demonstrated to be essential for pinpointing the role of T-cell infiltration and direct causal proteins (such as IDO1) in the expression of W > F substitutant peptides (Fig. 2) [13]. This shows that the expression of substitutant peptides is regulated by IDO1 expression, which is induced via the T-cell infiltration pathway.

Furthermore, ABPEPserver displays enrichment differences of substitutant peptides in tumours and tumour-adjacent normal tissues, an analysis that can be utilized for underpinning cancer-specific underlying mechanisms. Lastly, downloadable text files from the ABPEPserver can be used to identify common cancer-specific substitutants that can provide the foundation for predicting neoepitopes for a wide-ranging immunotherapeutic application. Altogether, ABPEPserver provides detailed information on the identity of the substitutants, their cancer-specific expression and immunotherapeutic potential.

Using the “W > F Substitutants” module for UCEC (Uterine Cancer) and the displayed scatter-plot, two example peptides ( fGHPAGK and SVLGCfK) were identified using fully substitutant database (database 1) and found to be expressed in a highly tumour-specific manner (73 tumours and 0 tumour-adjacent normal tissue, 45 tumours and 0 tumour-adjacent normal tissue respectively) (Fig. 3A). This tumour-specific expression implies that it is feasible to target these antigens specifically in cancerous tissues without harbouring any reactivity against normal cells if these peptides can present on the cell surface and bind to HLA molecules. Indeed, NETMHC [1] based prediction shows that many combinations of these two peptides have potentially strong binding affinity to one or multiple HLA super-alleles (Fig. 3B-C). This analysis indicates that the discovered peptides potentially harbour strong immunotherapeutic potential and warrant experimental validation. Thus, ABPEPserver can be used to identify potential cancer-specific antigens for immunotherapeutic applications.

In future, we plan to expand ABPEPserver functionality towards harbouring other kinds of aberrant peptides that are discovered, such as ribosomal-frameshift-associated chimeric peptides. In response to tryptophan shortage, it has been observed that ribosomes change the frame at the tryptophan-associated “TGG” codon, leading to the synthesis of W-chimera. Since W-chimera were only observed in cell-culture systems, it is important to analyze whether W-chimeras is also expressed in the eight cancer types analyzed here. If the expression is observed, the next pursuit is the association of gene expression pathways.