Enabling Variant Proteogenomic Assay Development Through a Peptide Target Selection Pipeline and Web Application Authors: Constantinos Blidjios, Pallab Bhowmick, Vincent R. Richard, Yassene Mohammed, Christoph H. Borchers Published: 2026 ABSTRACT ABSTRACTThe quantification of key protein variants can guide precision oncology to enable better identification of driver mutations. Bottom‐up assays infer parent protein levels through the quantification of surrogate peptides. Assay methods designed for absolute quantification of protein variants, such as multiple reaction monitoring mass spectrometry (MRM‐MS), can achieve quantitation with high precision and specificity. Bottom‐up approaches rely on proper selection of suitable surrogate proteogenotypic peptide targets to quantify protein variants of interest. To this end, we developed an R‐based bioinformatics pipeline to predict and evaluate variant‐specific peptides. The workflow generates variant protein sequences from wild‐type sequences and mutations encoded in the Human Genome Variation Society (HGVS) recommended nomenclature, performs in‐silico tryptic digestion, and identifies both variant and corresponding wild‐type peptides. Each peptide is evaluated by 37 selection criteria to determine suitability as MRM targets. To assess the strictness of these criteria, we applied all protein‐altering mutations from the NCI‐Genomic Data Commons (GDC) and COSMIC to our pipeline. Of the peptides outputted, 5% satisfied all defined criteria, representing the highest confidence candidates for assay development. We developed a database and web application from NCI‐GDC generated peptides for searching, filtering, and downloading. Another web application was developed to provide access to our pipeline. Significance Statement Targeted quantification of protein variants requires reliable selection of proteotypic peptides, yet this step remains a significant labor‐intensive stage of assay development. Our pipeline directly addresses this bottleneck by providing an automated, scalable system that converts any HGVS‐encoded mutation into rigorously evaluated variant‐specific peptide candidates. By integrating variant sequence generation, in‐silico digestion, and 37 stringent peptide quality criteria, the tool enables researchers to rapidly identify high‐confidence MRM targets. Large‐scale application of the pipeline against mutations from the NCI‐GDC and COSMIC, which catalogue variants observed in real human samples, further ensures robustness and broad applicability. The accompanying database (from NCI‐GDC generated peptides) and web applications make these resources immediately accessible, allowing users to search, filter, and download curated peptides or run the full pipeline on new mutations. This tool streamlines and accelerates an important first step in assay development, lowering the barrier to creating variant‐specific quantitative proteomics methods, which enables more reproducible and systematic measurement of clinically relevant protein alterations. Introduction In the field of quantitative proteomics, there is growing interest in determining the absolute levels of protein variants to supplement the methods for determining expression levels which are based on traditional genomics. Assessing protein variant levels in cancer tissues and biofluids can offer further insights into the status and progression of cancer [1, 2]. Understanding the mechanisms of cancer tumorigenesis requires analyzing the effects of mutations on cell regulation processes [3, 4, 5]. While current studies rely heavily on genomics approaches, proteogenomics can provide a more holistic view of cancer cell mutational progression [6, 7]. For example, it has been shown that mutants of human TP53 such as R175H promote upregulation of FOXH1, a transcription factor which can cause proliferation of cancer cells [8]. Another example is the mutation of IDH1, which results in neomorphic effects where the protein produces 2HG, a product which is thought to contribute to cancer metabolism [9]. In cases such as this, there would be benefits to supplementing such studies with proteomics information to determine whether the levels of variant being produced by such cells is sufficient to elicit the effects claimed. Thus, the development of quantitative proteogenomic assays to examine protein variant levels is warranted. Among the many methods currently used for protein quantification, mass spectrometry (MS) is a powerful technology that has gained popularity due to its reproducibility, and now, affordability [10, 11, 12, 13]. MS bottom‐up methods which incorporate internal peptide standards into the assay methodology can enable precise and “absolute” targeted proteogenomic assays, allowing researchers to quantify gene products resulting from specific mutations [14]. The selection of these target peptides, however, is a critical step since assay developers must consider a target's suitability to avoid wasting time and money on peptides which cannot be quantified. The criteria that researchers need to look for include a peptide's proteotypicity, stability, and ease of synthesis. Therefore, thorough background research on potential target peptides is essential. Suitable internal peptide standards used for bottom‐up quantitative proteomics can be labeled with heavy isotopes to create stable‐isotope labeled internal standard (SIS) peptides [15]. The use of SIS peptides in quantitation is important as it compensates for variations in MS instrument sensitivity, resulting in more consistent measurements across experiments, and enables absolute quantitation [15]. Peptide targets may also be selected based on other criteria established by researchers that indicate their suitability for quantification [16]. Selecting target peptides becomes challenging when specifically targeting protein variants, as there are limited databases that directly host variant sequences. Consequently, preliminary peptide target selection is both time‐consuming and cumbersome due to its manual nature and difficult to scale up when preparing to develop a large number of assays. This work aims to remedy this problem by developing a bioinformatics pipeline which can select and evaluate variant peptides from mutant protein sequences and evaluate the peptides for their suitability to be used in proteomics assays. Materials and Methods We developed our pipeline in R [17]. The pipeline has two main parts: one part for variant peptide generation and a second part for peptide evaluation and selection (Figure 1). We implemented the UniProt.ws, dplyr, httr, jsonlite, bioseq, and rentrez libraries to assist in the development of this pipeline [18, 19, 20, 21, 22, 23, 24]. The pipeline can generate preliminary peptide targets by generating variant protein sequences, digesting these sequences in‐silico with trypsin, selecting peptides containing the variant sequence and evaluating peptides by chosen selection criteria. Variant protein sequences are generated from wild‐type protein and cDNA sequences, and mutations encoded according to the Human Genome Variation Society (HGVS) nomenclature for proteins and cDNA (HGVSp, HGVSc) [25]. Protein and cDNA sequences were retrieved from EMBL‐EBI's Ensembl. A simplified outline of the peptide selection workflow. The user inputs a protein's accession number and mutation. The pipeline then automates peptide generation and evaluation for suitability as a target. The peptide evaluation procedure also retrieves additional information obtained from other knowledge bases. This includes information on protein canonical sequences, post‐translational modification (PTM) sites, the main functional chain of the protein, cleavage sites, additional annotations, eternal references, and the human proteome as retrieved from UniprotKB [26, 27]. Protein‐specific single nucleotide polymorphisms (SNPs) and population frequencies are retrieved from dbSNP [28]. Evidence on peptides previously observed in the literature is obtained from PeptideAtlas and gpmDB [29, 30]. Estimations of tryptic peptide digestion efficiency are calculated using EXPASY's PeptideCutter [31]. To evaluate the performance of our pipeline, we first selected all mutations from the National Cancer Institutes Genomic Data Commons (NCI‐GDC) and the Catalogue of Somatic Mutations in Cancer (COSMIC) which will result in a change at the protein level [32, 33]. These mutations were then applied to our pipeline and the time to process the output was measured. To evaluate the strictness of our pipeline, we counted the number of peptides from each database that adhered to all our selection criteria. Results and Discussion Peptide Evaluation Criteria To quantify proteins of interest by a bottom‐up approach, surrogate peptides selected to be targets must adhere to several criteria. These include being easy to synthesize, being a digestion product of the protein of interest unique to the peptidome, and having limited features which can negatively affect proteotypicity. Additionally, although our primary focus was on enabling variant peptide quantification, we also examined the wild‐type peptides, that is, the peptide which exists at the same region in the protein as the variant, but instead appears in the wild‐type protein. With this in mind, we implemented a total of 37 selection criteria for peptide suitability, which our pipeline will annotate, indicating that a peptide is suitable as a target for quantification purposes. Many of these selection criteria have been reported previously [16]. Target Peptide Ideal Composition In development of the pipeline, the suitability of the peptide for LC/MS was an important factor. For example, with SIS peptides which are synthesized de novo to include heavy‐labeled residues the length of the peptide is very important. The synthesis of longer peptides has higher costs with lower yields, and peptides which are too short may be difficult to detect by MS. Our pipeline therefore only considers peptides which are between 7 and 25 residues in length. Oxidizable peptides, such as those containing cysteine, methionine, and tryptophan residues, are also accounted for [34]. These residues can be problematic for quantification as oxidation changes the peptides' molecular weight, increasing the complexity of the MS spectrum and lowering the intensity of the signal for a specific molecular weight. Peptides containing other reactive motifs are problematic for the same reason. These include DG and DP, which can undergo hydrolysis; NG and QG, which can undergo deamination; N‐terminal glutamine residues, which can cyclize to form pyroglutamate [35]; and strings of repeating proline or serine, which can lead to deletions during synthesis [36]. As such, ideal target peptides are those which lack any of the aforementioned residues and motifs. Also important is the proteolytic digestion step which converts a protein sample into peptides. Peptides which have been selected for quantification need to be present in the sample mixture at a high enough abundance to have adequate detection. One factor affecting the abundance of a peptide product is the digestion efficiency of the protein, which is dependent on the enzyme selected and the residues flanking the cleavage sites. To predict peptide digestion efficiency, we implemented the model used in EXPASY's PeptideCutter tool [31]. We consider peptides to be ideal candidates if the product of digestion efficiencies of the cleavage sites at both the N and C‐termini of the selected peptide are above 10%. Target Peptide Uniqueness Peptides selected as targets for the quantification of protein variants should be unique to the digestion of the variant protein, and should never be observed in the wild‐type protein. We also checked that wild‐type peptide targets were only digestion products from the wild‐type protein and never from the variant. This ensures that quantifying the variant peptide means uniquely quantifying the variant protein. This also applies to the wild‐type. We also checked to ensure that the variant peptide sequence did not overlap with any other wild‐type protein in the UniProtKB human proteome. If these variant peptides were not unique, it would not be possible to ascertain which protein the peptide originated from, making it impossible to use as a target for quantification. Finally, wild‐type peptides were checked to see if they are present in all isoforms of the parent protein to ensure that the selected target is generalizable to all isoforms of the protein. Other Peptide Selection Criteria Given that clinical proteomic assays are typically applied to selected patient populations, it is important to assess the generalizability of the candidate peptide targets. To address this, we evaluated whether the wild‐type peptide sequences overlapped with canonical protein sequences as defined by UniProtKB. Canonical sequences generally represent the most functionally relevant and widely expressed isoforms of a given protein [37]. Verifying that peptide targets align with canonical sequences also provides an estimate of their potential applicability across diverse populations. Furthermore, this approach enables the extrapolation of annotation such as PTMs and single‐nucleotide polymorphisms from canonical to non‐canonical sequences that share conserved motifs. Target peptides ideally should not contain any significant SNPs—that is, variations that can occur with different frequencies in different populations but which are not considered to be mutations. This is also important for ascertaining the generalizability/applicability of a peptide target, as some SNPs are highly prevalent in certain populations. We used the dbSNP resource for information on SNP frequency [28], and we have set a cut‐off of greater than 1% presence in the population for significance in our pipeline. Reliably predicting peptide proteotypicity can present challenges. For this, we referred to the PeptideAtlas and the gpmDB, knowledge bases which host information on whether or not a peptide has been previously observed in the literature [29, 30]. Having direct evidence that a peptide has been observed previously is important, although this information is usually limited to wild‐type peptides. Peptides containing PTM sites are not ideal due to synthesis difficulties and the possibility of a sample of peptides being differentially modified, making accurate quantification difficult. Additionally, PTMs on residues flanking a cleavage site may impede digestion. Peptide targets containing cleavable regions are not ideal since proteins in a sample may be cleaved at this site. This means that the digestion of such a protein might not yield the target peptide, making quantification difficult based on this peptide. Target peptides preferentially originate from the functional main chain of the protein. Signal peptides, for example, are not considered to be part of a protein's main functional chain because signal peptides are attached to newly synthesized proteins and are generally used for protein translocation and they are cleaved off afterwards. Peptide Database Using NCI‐GDC Mutations To evaluate the peptide selection pipeline, mutations were retrieved from the online knowledge bases hosting human patient data, such as the NCI‐GDC and COSMIC knowledgebases, and these mutations were subsequently used in the pipeline. To preface this discussion, we highlight that the large majority of mutations analyzed will result in only a single variant peptide being produced. For mutations which can result in many variant peptides, the pipeline only includes the first variant peptide generated in the sequence. This allows us to compare 1:1 the number of mutations to the number of peptides produced. Additionally, our pipeline accepts mutations of all types as input but evaluates only those which result in a change in the protein sequence. Therefore, any mutations in gene regulatory regions or intronic regions are not considered. The pipeline does, however, evaluate all mutation types which do result in protein sequence changes. This includes substitutions, deletions, duplications, insertions, deletion‐insertions, frameshifts, and extensions—all which can be used in the pipeline via HGVS nomenclature. See Figure 2 for a breakdown of peptides generated by considering all mutant peptides derived from the NCI‐GDC and COSMIC knowledgebases. Each qualifying mutation obtained from the NCI‐GDC or COSMIC was used to generate a single variant peptide. This enables a comparison of the number of mutations included to the number of variant peptides generated. Here we show the total number of mutations from the NCI‐GDC and COSMIC knowledgebases used in our pipeline. We also show the number of variant peptides generated as a consequence of these mutations, and how many of these peptides pass all of the selection criteria, indicating their suitability for use as target peptides. We observed that ~5% of all protein‐level variants will result in a peptide which adheres to all of the defined selection criteria. This number is a result of both the strictness of these selected criteria, and the enzyme used for digestion. Figure S1 shows which selection criteria resulted in the most rejections of a peptide candidate—that is, which selection criteria are the most stringent. From Figure S1, we can see that the most prevalent cause for rejection of a peptide candidate is the length of the peptide. This is an unavoidable consequence of using trypsin as the digestion enzyme, as a significant portion of the generated tryptic peptides are outside of this range (Figure 3A,B). Variants which cannot be studied using tryptic peptides could potentially see better success using a different enzyme, such as chymotrypsin. For some variants, doing so may result in peptides that have more desirable characteristics for use as a target, but this will depend on the protein sequence. (A,B) Shows the distribution of variant peptide sequence lengths for peptides generated by the pipeline using NCI‐GDC and COSMIC mutations. (C,D) Shows the distribution of mutation types from NCI‐GDC and COSMIC used to generate peptides using our pipeline. Other notably high criteria resulting in peptide rejection include the presence of oxidizable residues in the peptide of interest, lack of uniqueness in the peptidome, and lack of pre‐reported evidence of the peptides existence in other peptide databases. Figure 3C,D shows that the large majority of variant proteins examined contain only single amino acid changes, meaning that among the protein's digestion products, only one peptide would be a variant peptide. The consequence of this is that options are limited for potential targets using this method. A solution to this problem would be to use different proteolytic enzymes which may yield better surrogate peptides for certain mutations/variant proteins. As our pipeline currently only evaluates tryptic peptides, expanding the number of enzymes is an interesting possibility. In a cross‐comparison of the NCI‐GDC and COSMIC mutation datasets, we observed an overlap of 635,529 mutants between the two. Lastly, we also benchmarked the pipeline by taking 1000 mutations from the NCI‐GDC and measuring the time to generate the output. The measured time to process was 6.4 min on average. Web Applications Hosting the Peptide Selection Pipeline and Database To disseminate the software which we developed during this study, we have written two web applications: one implementing the peptide‐generation pipeline and another providing a searchable database of peptides and annotations derived from the NCI‐GDC mutations. The peptide‐generation pipeline application features a simple interface in which users can upload desired protein accession numbers and mutations of interest through our provided template. The output is made available as a CSV file for download and includes the generated peptides, and the selection criteria. The tool allows the end user to select any desired combination of selection criteria to filter their resulting output list. Table 1 shows the full list of available filters and recommended hierarchy of importance. The importance of certain criteria may vary depending on the field of study and the availability of alternative methodologies, such as obtaining peptide standards through approaches other than chemical synthesis (e.g., SILAC mouse) [38]. This deployed application named QuaVaPeptidePicker can be accessed at https://quavapeptidepicker.proteomicscentre.com/. The searchable database application supports queries based on proteins, peptides, mutations, and other relevant attributes. Search results can be filtered by the selection criteria, and downloaded. We also include hyperlinks to all external resources accessed for simplified cross‐referencing. Visual graphics of the peptide's position within its parent protein is also available. This deployed application named QuaVaProt can be accessed at https://quavaprot.proteomicscentre.com/. All source code related to the development of the pipeline and web applications are publicly available at https://github.com/qualisomics/QuaVaProt. TABLE 1List of peptide selection criteria with recommended hierarchy of importance.Criteria labelCriteria pass conditionImportanceReasoningCanonical checkWT peptide overlaps with canonical sequenceCriticalRequired for quantificationVariant unique(vs WT)Variant peptide is not part of the WT proteinCriticalRequired for quantificationWT unique(vs variant)WT peptide is not part of the Variant proteinCriticalRequired for quantificationVariant unique in ProteomeVariant peptide is not part of any other WT protein in proteomeCriticalRequired for quantificationWT unique in proteomeWT peptide is not part of any other WT protein in proteomeCriticalRequired for quantificationVariant DGVariant peptide does not contain Asp‐GlyHighAvoids differentially modified peptides (hydrolysis)Variant DPVariant peptide does not contain Asp‐ProHighAvoids differentially modified peptides (hydrolysis)Variant NGVariant peptide does not contain Asn‐GlyHighAvoids differentially modified peptides(deamination)Variant QGVariant peptide does not contain Gln‐GlyHighAvoids differentially modified peptides(deamination)Variant PPPVariant peptide does not contain Pro‐Pro‐ProHighIdeal for chemical synthesis of standardVariant PPGVariant peptide does not contain Pro‐Pro‐GlyHighIdeal for chemical synthesis of standardVariant serine stringVariant peptide does not contain multiple sequential serinesHighIdeal for chemical synthesis of standardWT DGWT peptide does not contain Asp‐GlyHighAvoids differentially modified peptides (hydrolysis)WT DPWT peptide does not contain Asp‐ProHighAvoids differentially modified peptides (hydrolysis)WT NGWT peptide does not contain Asn‐GlyHighAvoids differentially modified peptides(deamination)WT QGWT peptide does not contain Gln‐GlyHighAvoids differentially modified peptides(deamination)WT PPPWT peptide does not contain Pro‐Pro‐ProHighIdeal for chemical synthesis of standardWT PPGWT peptide does not contain Pro‐Pro‐GlyHighIdeal for chemical synthesis of standardWT serine stringWT peptide does not contain multiple sequential serinesHighIdeal for chemical synthesis of standardIsoform checkWT peptide is part of all isoforms of the proteinHighIdeal for generalizabilityPTMWT peptide does not contain a PTM site, nor on flanking residuesHighAvoids differentially modified peptides, avoids potential peptide missed cleavageCleave siteWT peptide does not contain a cleavage siteHighAvoids differentially modified peptides(cleavages)SNPWT peptide does not contain a SNP with > 1% prevalenceHighIdeal for generalizabilityPeptide existsWT peptide has been previously reported in literatureHighIdeal for quantificationVariant digestion efficiencyVariant peptide has a predicted digestion efficiency of > 10%HighIdeal for quantificationWT digestion efficiencyWT peptide has a predicted digestion efficiency of > 10%HighIdeal for quantificationVariant lengthVariant peptide length is 7–25 residuesLowIdeal for chemical synthesis of standardWT lengthWT peptide length is 7–25 residuesLowIdeal for chemical synthesis of standardVariant N GlnVariant peptide does not contain N‐terminal GlnLowAvoids differentially modified peptides(cyclization)WT N GlnWT peptide does not contain N‐terminal GlnLowAvoids differentially modified peptides(cyclization)Variant CVariant peptide does not contain CysLowAvoids differentially modified peptides(oxidation)Variant MVariant peptide does not contain MetLowAvoids differentially modified peptides(oxidation)Variant WVariant peptide does not contain TrpLowAvoids differentially modified peptides(oxidation)WT CWT peptide does not contain CysLowAvoids differentially modified peptides(oxidation)WT MWT peptide does not contain MetLowAvoids differentially modified peptides(oxidation)WT WWT peptide does not contain TrpLowAvoids differentially modified peptides(oxidation)In main chainWT peptide is categorized as part of protein main chainLowPrimarily a flagJohn Wiley & Sons, Ltd. List of peptide selection criteria with recommended hierarchy of importance. Variant Peptide Picking Tool Test Case An example test case is provided here to demonstrate the functionality of QuaVaPeptidePicker. The input will comprise 100 mutations obtained from the NCI‐GDC, where a standardized template (obtained from the website) is populated with the Ensembl transcript accession number, UniprotKB accession number, HGVSc, HGVSp, and mutational consequence label for each variant. It is essential that the structure of the template remains unaltered, such that no columns are deleted or renamed, and that HGVSp annotations are reported using single‐letter amino acid notation. Detailed formatting instructions are provided in the “Help” section of the QuaVaPeptidePicker website. Upon uploading the completed template, users are presented with a preview of the full dataset in a confirmation dialog prior to submission. After confirmation, the tool will begin processing the upload. For this test case, the total processing time was 1 min 25 s. Upon completions of the operation, the website displays a preview table with a subset of the output fields, including the peptides generated by the tool. Users may then apply desired selection criteria to filter the results. Downloading the resulting CSV file provides access to the complete information retrieved/generated for each variant, including all selection criteria. For this test case, application of all available filters to our output reduced the number of resulting entries for 100 to nine final candidates. The complete input dataset is provided as Table S1, and the corresponding output is provided as Table S2. Conclusion In summary, in this paper we present a pipeline designed to streamline the development of proteogenomic assays, with a particular focus on the selection of target peptides for the quantification of protein variants. By automating the peptide target selection process and integrating supporting data from external references, the pipeline significantly accelerates assay development while reducing manual effort. The accompanying web application offers an accessible, user‐friendly interface that allows researchers to customize peptide selection criteria according to their experimental needs. Future development will focus on expanding enzymatic options to further enhance the pipeline's flexibility and utility. By making the software open‐source and freely accessible to the research community, we aim to support broader adoption and to facilitate continued advancements in proteogenomic research. Conflicts of Interest Christoph H. Borchers is the founder and Scientific Advisor of MRM Proteomics, Inc., and the VP of Proteomics at Molecular You. The other authors declare no conflicts of interest. Supporting information Supporting File 1: pmic70139‐sup‐0001‐FigureS1.pdf. Supporting File 2: pmic70139‐sup‐0002‐TableS1.csv. Supporting File 3: pmic70139‐sup‐0003‐TableS2.csv.