Seattle, WA
22 - 23 July, 2024

The 2023 Program

Monday, July 17
Tuesday, July 18
Registration & Continental Breakfast
Chair: Rob Moritz (ISB)
Introductions by Platinum Sponsors
Session 1: Proteomics Applications to Disease
Chair: Chris Overall (UBC)
Philipp Lange (UBC)
DIA proteome guided therapies in precision oncology
GD Barnabas, TA Bhat, V Goebeler, CA Maxwell, GS Reid, DL Senger, S Parker, N Azzam, JN Berman, J Bush, C Strahlendorf, R Deyell#, CJ Lim#, PF Lange# on behalf of the PROFYLE program

1.Department of Pathology, University of British Columbia, Vancouver, BC, Canada 2.Department of Pediatrics, University of British Columbia, Vancouver, BC, Canada 3.Michael Cuccione Childhood Cancer Research Program, BC Children’s Hospital and Research Institute, Vancouver, BC, Canada 4.Department of Biochemistry, University of British Columbia, Vancouver, BC, Canada 5.Department of Medicine, McGill, Montreal, QC, Canada 6.Lady Davis Institute for Medical Research, Jewish General Hospital, Montreal, QC, Canada

Precision treatments that target molecular alterations in cancer have significant potential to improve therapy options for hard-to-treat cancers and reduce late effects in general. Genome sequencing has laid the foundation for precision medicine, yet, clinical success remains moderate. Identifying therapeutic targets at the protein and pathway levels holds great promise.
Here, we present the case of an adolescent with metastatic, progressive spindle epithelial tumor with thymus-like differentiation (SETTLE) and evaluate how quantitative proteome profiling can identify treatment options not apparent at the genome or transcript level. DIA proteome analysis of macro-dissected tumor and adjacent normal from formalin-fixed paraffin-embedded sections was completed within two weeks of biopsy and identified key proteins as possible targets for single or combination therapy. We validated evlevated abundance of the target by immunohistochemistry in comparison to levels across AYA tumors and confirmed positive drug response in vitro and in patient-derived model systems. Following failure of cytotoxic chemotherapy and second-line sorafenib treatment, a monotherapy trial was initiated by the patient but stopped after 8 weeks after evidence of a moderate reduction in growth rate but overall progressive disease. Possible combination therapies were evaluated further in the patient-derived models.
We show that DIA proteome analysis of FFPE biopsies can identify treatment targets within two weeks and inform patient care in a clinically meaningful timeframe. Yet, timing remains a serious challenge to precision medicine and in particular in the context of combination therapies. Hypothesizing that a change in personalized pediatric oncology from the current reactive to a proactive approach initiated at diagnosis is feasible if actionable genetic lesions and protein pathways are stable throughout disease progression, we will present data showing genome and proteome stability in B-ALL supporting that precision diagnostics already be conducted at initial diagnosis.
Wei-Jun Qian (PNNL)
Spatial Proteomics Profiling of Intra-donor Pancreatic Islet Heterogeneity in Prediabetic Subjects Reveals an Immune Signature of Progressive Islet dysfunction
Ying Zhu1, An D. Fu1, Sarah Williams1, Elizabeth A. Butterworth2, Tyler Sagendorf1, Yumi Kwon1, Adam C Swensen1, Clayton E Mathews2, Martha Campbell-Thompson2, and Wei-Jun Qian1

1Biological Sciences Division and Environmental Molecular Sciences Laboratory, Pacific Northwest National Laboratory, Richland, Washington 99352, 2Department of Pathology, Immunology, and Laboratory Medicine, University of Florida, Gainesville, FL 32610.

PURPOSE: Type 1 diabetes (T1D) is a complex autoimmune disease, and its clinical diagnosis is preceded by an asymptomatic phase where autoimmunity and progressive β-cell dysfunction occur. As it is not ethical to obtain longitudinal human pancreas tissue samples, recent advances in single cell ‘omics open novel opportunities to study the “pseudo-time” of disease progression by analyzing cellular heterogeneity using samples from organ donors with pre-diabetes. Herein, we explore the intra-donor islet heterogeneity in presymptomatic, autoantibody positive (AAb+) donors by applying single islet proteomics with the goal of revealing a detailed mechanistic understanding of early β-cell dysfunction in T1D.
METHODS: Human fresh frozen islet sections from presymptomatic, multiple AAb+ cases from nPOD program were isolated by laser microdissection (LMD) following islet phenotyping on adjacent sections by immunofluorescence for insulin and CD3.
Each islet section was collected into separate wells for single islet proteomics analysis. 102 and 104 islet sections were collected from two multiple AAb+ donors (nPOD case IDs 6450 and 6521), respectively.
Single islet proteomics based on NanoPOTS-LC-MS/MS technology was performed on all islet sections to explore intra-donor islet heterogeneity on a Thermo Scientific Orbitrap Fusion Lumos Tribrid Mass Spectrometer. Label-free quantification data were analyzed using the Frag-Pipe package.
RESULTS: To identify the molecular determinants of early β-cell inflammation and stress, we have conducted highly sensitive single islet proteomics analysis of ~100 islet sections per donor isolated by LMD from two multiple AAb+ donors with islet subpopulations phenotyped by immunofluorescence. The large-scale single islet proteomics analyses using label-free quantification resulted in quantification of 4500-5000 proteins across all islet samples. UMAP (Uniform Manifold Approximation and Projection) clustering analysis reveals multiple islet clusters for each donor with several islet clusters displaying loss of pancreatic beta cell and delta cell markers, indicative of beta cell dysfunction. Moreover, the weighted correlation network analysis (WGCNA) clearly identifies an immune-related protein cluster (immune signature), which correlates the CD3 phenotyping data. Based on this “immune signature”, the single islets from each donor can be clustered into three “subtypes” of immune-low, immune-moderate, immune-high, respectively, reflecting a “pseudo-time” of progression to beta cell dysfunction in presymptomatic cases. While this work is still preliminary in that two donors were profiled at the single islet levels, it is noteworthy that the patterns of ‘immune signature” and islet subtyping are consistent between both donors. This work demonstrates the potential of in situ near-single-cell proteomics for exploring cellular heterogeneity to obtain mechanistic information related to disease progression.
Further work in validating the discovery results in additional mAAb+ donors is still ongoing in conjunction with matched control donors.
CONCLUSION: Single islet spatial proteomics analysis of intra-donor heterogeneity in presymptomatic AAb+ donors revealed interesting molecular level information related to the early progression of autoimmune mediated islet dysfunction in human T1D.
Deanna Plubell (UW)
Brain derived peptide abundance correlations reveal molecular subtypes of sporadic Alzheimer’s disease.
Deanna Plubell, Genn Merrihew, Jea Park, Chris Hsu, Christine Wu, Thomas Montine, Michael MacCoss

Dept of Genome Sciences, University of Washington; Dept of Pathology, Stanford University

Alzheimer’s is a complex disease that occurs with or without causative mutations and can be accompanied by a range of age-related comorbidities. This diverse presentation makes it difficult to study molecular changes specific to AD. To study the molecular signatures of disease we constructed a unique human brain sample cohort inclusive of autosomal dominant AD (ADD), sporadic ADD, and those without dementia with high AD histopathologic burden, and cognitively normal individuals with no/minimal AD histopathologic burden. Using DIA-MS we examined differences between groups at a peptide and protein level in four brain regions. Cognitive status was determined by DSM-IVR criteria; and neuropathologic change was determined by consensus diagnosis. Tissue was obtained from hippocampus, superior and medial temporal gyri, inferior parietal lobe, and caudate nucleus brain regions. For each of the four regions samples were divided into batches which included both inter-batch external quality controls and inter-brain region quality controls. A one µg aliquot of each digested sample was separated by nHPLC-MS/MS on an Easy nanoLC and analyzed by DIA on with Orbitrap Lumos. Peptide detection and signal processing was performed with EncyclopeDIA using DIA libraries and predicted human spectral libraries. Qualitative and quantitative metrics were assessed using Skyline and EncyclopeDIA. Across brain regions protein group detections ranged from 4504 in the caudate nucleus (CN) to 6031 in the inferior parietal lobe (IPL); while peptide detections ranged from 26135 in CN to 42500 in IPL. By comparing peptide measurements against known pathological markers of disease we found that a subset of peptides separates sporadic AD into at least two subgroups. Across all brain regions, peptides derived from the amyloid-β sequence are most abundant in autosomal dominant AD, followed by sporadic AD, high neuropathological, and low neuropathological controls. Since these peptide abundances reflect the diagnosis of each group, they were used to find peptides with correlated abundance profiles. For example, in the SMTG we find peptides from 165 proteins are positively correlated with amyloid-β abundances, while peptides from 157 proteins are negatively correlated. Interestingly, the peptides correlated with amyloid-β peptide abundance separate the sporadic AD patient samples into two distinct subgroups by hierarchical clustering and by principal component analysis. One of the subgroups of Sporadic have profiles that more closely resemble those of autosomal dominant disease, and they are a younger average age compared to the rest of the sporadic cases. In addition to these differences being associated with age, they also appear to be related to the severity of disease, as indicated by the last MMSE score and tau peptide abundance differences. When treated as separate groups - i.e., ‘early’ and ‘late’ sporadic AD, a large number of peptides are significantly different between them in the SMTG. This difference between sporadic subgroup signatures is not as strong in other brain regions, with fewer significantly different peptides in the IPL and none in the caudate nucleus. Overall, the SMTG and IPL share some similarities, but have many unique differences across the experimental groups. The caudate nucleus is very different from both the IPL and SMTG, instead having the most significant differences found between the autosomal dominant and early sporadic subtype, and between the two non-dementia control groups. These results indicate that the signature of disease varies across brain regions, and also varies within sporadic AD.
Chuwei Lin (UW)
HDAC inhibitors induce proteome remodeling of diverse cancer cells
Chuwei Lin, Devin K Schweppe

Genome Sciences, University of Washington

Lung cancer is a significant global public health concern and is the leading cause of cancer-related deaths. Small-molecule inhibitors have been the main therapeutic drugs for lung cancer. Over the past few decades, epigenetic mechanism has emerged as a crucial player in cancer initiation and progression. Histone acetylation process is a main process in epigenetics. Histone deacetylases (HDACs) are enzymes that remove acetyl group from histone, thereby reducing gene transcription. Aberrant expression of HDACs is linked to human cancers, making them potential targets for therapeutic intervention. Among them, HDAC1, 2, and 3 have been observed to have increased expression in lung cancers. HDAC inhibitors (HDACi) are a group of epigenetic drugs which inhibit the activity of one or multiple HDACs. HDACi has been shown to induce cell cycle arrest, differentiation, and programming cell death, exhibiting powerful anticancer activities. However, cellular responses to HDAC inhibition are heterogenous and dependent on factors such as the genetic background, metabolic state, and on-/off-target engagement of individual HDACi drugs, which has been largely overlooked in previous research. This project aims to collect and characterize lung cancer cell lines derived from different genetic backgrounds and then profile the proteome of different cell lines in response to different HDACi. By quantifying proteome remodeling of HDACi-treated cell lines with diverse genetic backgrounds, we can determine how diverse cellular factors drive heterogeneous responses towards HDACi.
Lightning Talk: Jim Sanford (PNNL)
Proteomic and phosphoproteomic responses to acute and chronic endurance exercise in rats
James Sanford1, Tyler Sagendorf1, Gina Many1, Chelsea Hutchinson1, Marina Gritsenko1, Hugh Mitchell1, James Pino1, Laurie Goodyear2, Sue Bodine3, Paul Piehowski1, Wei-Jun Qian1, Joshua Adkins1, and The MoTrPAC Study Group

1Pacific Northwest National Laboratory, Richland, WA 2Joslin Diabetes Center, Harvard Medical School, Boston, MA 3Oklahoma Medical Research Foundation, Oklahoma City, OK

Though the health benefits of physical activity are widely appreciated, the molecular mechanisms affected in response to exercise are not completely understood. Through the Molecular Transducers of Physical Activity Consortium (MoTrPAC), we sought to characterize the proteomic and phosphoproteomic responses in multiple tissue types from models of acute and chronic endurance exercise in six-month-old rats. Utilizing multiplexed isobaric labeling and offline fractionation coupled with LC-MS/MS analysis, we obtained some of the deepest reported coverage of the proteome and phosphoproteome of skeletal muscle, white adipose tissue, lung, and kidney—quantifying greater than 10,000 unique proteins and 30,000 unique phosphorylation sites in most tissues. Following a single bout of treadmill running, substantial changes to the phosphoproteome were observed at early timepoints (immediate post-exercise to 1-hour post-exercise) in skeletal muscle and lung, which largely returned to baseline by 24 hours. Global protein abundance changes increased over time in all tissues, with the largest number of significant changes observed at 24 hours post-exercise in proteins related to tissue organization, stress responses, and metabolic processes. Endurance exercise training for up to 8 weeks led to robust changes in the proteome of skeletal muscle and white adipose tissue, as well as alterations in the phosphoproteome that suggest differential activity of kinases involved in systemic metabolism. In all tissues profiled, we observed strong sex differences in baseline protein abundance, with several tissues showing sexually dimorphic exercise-induced proteomic and phosphoproteomic responses. Together with additional multi-omic measurements through other MoTrPAC centers, these data provide a tremendous resource for advancing our understanding of the molecular impacts of physical activity.
Lightning Talk: Jesse Trejo (PNNL)
Optimizing proteomic analysis of liquid-based Pap tests for biomarker discovery studies
Jesse B. Trejo, Kristin L.M. Boylan, Ashley J. Petersen, Joshua R. Hansen, Tao Liu, Amy P.N. Skubitz, Paul D. Piehowski

PNNL, University of Minnesota

The early detection of ovarian cancer is important to improve treatment outcomes for patients, but an effective screening for diagnoses has not been developed yet. With our collaborators at the University of Minnesota, we are investigating liquid-based Pap samples as a potential source for ovarian cancer biomarkers. However, proteomic analysis of these samples presents two major challenges: shed proteins are highly dilute within the fixative solution, and standard storage of this fixative has introduced the interference of polyethylene glycol (PEG) contamination. Two different processing methods have been optimized for ‘Mock’ SurePath Pap tests and ThinPrep Pap tests to tackle these challenges and are presented here. The first method consists of an acetone precipitation to remove PEG contamination followed by resolubilization in 8M Urea for a standard digestion protocol. We found that SurePath samples yielded enough protein material for discovery proteomics with Tandem Mass Tag (TMT) labeling. Furthermore, we found that adding a 70% ethanol wash after the acetone precipitation improved reliability to remove PEG contamination without negatively impacting protein yield. A downside of the acetone precipitation is the resulting loss of protein, which is a problem when working with small samples such as the ThinPrep Pap samples. Applying the above protocol resulted in protein yields below the detectable limit in a BCA assay, and the lack of a visible pellet made PEG removal unreliable. This caused inconsistent results and <500 protein identifications per sample even when we pooled multiple samples together. To address this, we applied the S-Trap Micro digestion protocol which is specifically intended for low yield samples. PEG contamination was completely removed from the samples and protein identifications from individual samples increased 3-fold. With these optimizations, we have carried out discovery proteomics studies using a cohort of patients with SurePath Pap tests to identify several putative biomarker candidates for downstream validation using targeted proteomics. Proteomic analysis of the ThinPrep cohort is currently underway and we plan to analyze larger cohorts in the future to continue the identification of biomarkers in ovarian cancer.
Lightning Talk: Gennifer Merrihew (UW)
A systematic method to quantify peptides in CSF for the analysis of neurodegenerative diseases
Gennifer Merrihew1; Jea Park1; Deanna Plubell1; Julia Robbins1; Brian Searle2; Eric Huang1; Christine Wu1; Kathleen Poston2; Thomas Montine2; Michael J. MacCoss1

1University of Washington, Seattle, WA; 2Ohio State University, Columbus, OH; 3Stanford University, Stanford, CA

Neurodegenerative diseases have complicated comorbidities and mixed pathologies that are difficult to diagnose. To assess biological signatures of neurodegenerative diseases, we used a systematic method to collect quantitative proteomics data from an age and sex matched, clinically defined cohort of cerebrospinal fluid (CSF) from 280 patients categorized into four groups: healthy control, Alzheimer’s disease/mild cognitive impairment, Parkinson’s disease cognitively normal and Parkinson’s disease cognitively impaired. To confidently measure the same peptide quantities from different patient samples, we developed a series of controls to assess system suitability, individual sample preparation quality, and quantitative accuracy on the batch level. These controls were tracked in real-time using Skyline, AutoQC, and PanoramaWeb.
CSF samples were resuspended in SDS lysis buffer with yeast enolase protein to assess sample digestion and then reduced and alkylated. Proteins were aggregated on MagResyn Hydroxyl beads, washed, and digested to peptides, using a Thermo KingFisher Flex. Peptides were separated using reverse-phase chromatography with a Thermo Easy nano-LC and electrosprayed into a Thermo Orbitrap Eclipse Tribrid analyzed using a 12 m/z staggered DIA isolation scheme. The data were demultiplexed to 6 m/z with Proteowizard.
EncyclopeDIA was used to detect peptides and assign peak boundaries which were then imported into Skyline. Peptide quantities were then median normalized and batch corrected. All levels of data are available via PanoramaWeb.
Lumbar CSF from 280 patients were divided into three major groups: 1) Healthy Control, 2) Alzheimer’s disease/mild cognitive impairment, and 3) Parkinson’s disease cognitively normal and impaired. Each row of half a 96-well plate contained 10 balanced and randomized CSF samples and two external controls. One external inter-batch control was a pool of 50 patients representing all 4 groups and the other external inter-experiment control was a commercially available pool of CSF. These controls were used to assess the technical precision within and between each batch prior to and following normalization and batch adjustment. The coefficient of variation calculated for all peptides in the inter-batch control improved 7.75% after normalization and batch adjustment at the peptide-level and 20.38% at the protein level. We also use internal process controls to assess sample preparation and monitor data collection. These controls were necessary, as the data collected encompassed seven batches analyzed on separate LC columns and traps over a three-month period.
Preliminary analysis of the CSF from the 280 patients yields 23,742 peptides and 2685 protein groups. The peptides found in our data represent many of the expected proteins previously shown to be associated with neurodegenerative diseases but we also find new proteins.
Our data has also revealed statistical significant positive or negative peptide correlations with known Alzheimer’s disease biomarkers amyloid-beta 42/40 and pTau181. Pairwise group comparisons using WGCNA yield gene ontologies such as carbon metabolism, apoptosis, neutrophil degranulation, focal adhesion and neurodegeneration.
Session 2: Computational Proteomics
Chair: Eric Deutsch (ISB)
William Noble (UW)
CONGA: Combining open and narrow searches with group-wise analysis
Jack Freestone, William Noble, Uri Keich

University of Sydney, University of Washington

Mass spectrometry is limited in its ability to detect post-translationally modified peptides due to the associated large search space. Open modification searching addresses this problem by comparing spectra against peptides whose masses can differ significantly from the associated precursor mass. Comparing open- with traditional "narrow-window" search, we found that open searches on their own often produce fewer discovered peptides than corresponding narrow searches. Only when coupled with post-processors such as Percolator or PeptideProphet do open searches typically become better than narrow searches. However, our analysis showed that such post-processors can fail to control the false discovery rate (FDR). This observation motivates our CONGA method, which combines results from narrow and open searches, improving on both while rigorously controlling the FDR.
Will Fondrie (Talus)
Dive Deeper with Depthcharge: A Transformer Toolkit for Mass Spectrometry Data
William E Fondrie (1), Wout Bittremieux (2), Melih Yilmaz (3), William S Noble (3,4)

(1) Talus Bioscience, Seattle, WA (2) University of Antwerp, Antwerpen, Belgium (3) Paul G Allen School of Computer Science and Engineering, Seattle, WA (4) Department of Genome Sciences, University of Washington, Seattle, WA

Introduction Deep learning has revolutionized the analysis of mass spectra; from predicting the tandem mass spectrum generated by an analyte, to sequencing peptides from mass spectra de novo, the neural network models that underpin deep learning are now ubiquitous. In recent years, a neural network architecture called the transformer has become the architecture of choice for developing state-of-the-art deep learning models, in domains including natural language processing, protein structure prediction, and importantly, the analysis of mass spectra. However, every new model developed for mass spectra has essentially been forced to start from scratch. Here, we introduce depthcharge, an open-source deep learning framework that provides the building blocks for transformer models of mass spectra and the analytes that generate them.
Methods A tandem mass spectrum can be described as a bag of peaks where each peak is defined as a pair of m/z and intensity values. The distances between m/z values, the m/z values themselves, and their associated intensities provide structural information about the analyte; hence, we hypothesize that the self-attention mechanism which characterizes the transformer architecture would be ideal for learning the relationships among peaks within a mass spectrum, similar to the relationships among words within a sentence. Additionally, peptides and small molecules can be represented as sequences of tokens (either a peptide sequence or SMILES string). Depthcharge provides PyTorch modules to parse, batch, and encode these data structures and use them to build transformer models.
Preliminary Data Depthcharge provides the building blocks to build transformer models for mass spectra and common analytes, such as peptides and small molecules. Unlike other previous architectures, such as recurrent neural networks, transformers lack a built-in representation for the order of elements in the input sequence; position in the sequences is generally encoded as a sequence of sinusoids that is summed with a representation of each element. We use this quality of transformers to our advantage to model mass spectra: the m/z values are encoded as a series of sinusoids and summed with a learned representation of the intensity. We illustrate both how this process takes place and demonstrate that this method provides a high fidelity representation of a mass spectrum.
We then present a series of case studies on the various ways that depthcharge can be used, demonstrating the configurations required for, predicting peptide properties such as collisional cross section, predicting the b and y ion intensities generated from a peptide precursor, and co-embedding peptides and mass spectral into the same latent space. In each case, we build a minimal model atop depthcharge and outline the components required to build it. We then compare each against current tools in the field, demonstrating that even these minimal models are capable of achieving high-quality results. Finally, we show that these models require relatively few lines of code to implement due to the tools provided by depthcharge.
We aim for depthcharge to provide a user-friendly, foundational framework that will propel biological discovery through new models of mass spectrometry data. Depthcharge is open-source and available under the permissive Apache 2.0 license:
Michael Hoopmann (ISB)
Ving: A New Tool in the Trans-Proteomic Pipeline for XL-MS Using Cleavable Cross-Linking Reagents
Michael R. Hoopmann, David D. Shteynberg, Luis Mendoza, Kamal Mandal, Arun P. Wiita, Eric W. Deutsch, Robert L. Moritz

Institute for Systems Biology, University of California, San Francisco

Mass spectrometry-cleavable chemical cross-linkers fragment in the gas phase using collision induced dissociation at low energy levels. Combined with MS3 acquisition methods, two cross-linked peptides are separated, then independently selected and analyzed. Using this approach, the computational burden of identifying both peptides in the same spectrum is eliminated, enabling existing database search tools to identify cross linked peptides. Here we present Ving, a new tool that extends cleavable cross-linking analysis capabilities to the Trans Proteomic Pipeline.
Lightning Talk: Lincoln Harris (UW)
Evaluating proteomics imputation methods with improved criteria
Lincoln Harris, William E. Fondrie, Sewoong Oh, William S. Noble

University of Washington, Department of Genome Sciences Paul G. Allen School of Computer Science and Engineering, University of Washington Talus Biosciences

Quantitative measurements produced by tandem mass spectrometry proteomics experiments typically contain a large proportion of missing values. This missingness hinders reproducibility, reduces statistical power, and makes it difficult to compare across samples or experiments. Although many methods exist for imputing missing values in proteomics data, in practice, the most commonly used methods are among the worst performing. Furthermore, previous benchmarking studies have focused on relatively simple measurements of error, such as the mean-squared error between the imputed and the held-out observed values. Here we evaluate the performance of a set of commonly used imputation methods using three practical, “downstream-centric” criteria, which measure the ability of imputation methods to reconstruct differentially expressed peptides, identify new quantitative peptides, and improve peptide lower limit of quantification. Our evaluation spans several experiment types and acquisition strategies, including data-dependent and data-independent acquisition. We find that imputation does not necessarily improve the ability to identify differentially expressed peptides, but that it can identify new quantitative peptides and improve peptide lower limit of quantification. We find that MissForest is generally the best performing method per our downstream-centric criteria. We also argue that existing imputation methods do not properly account for the variance of peptide quantifications and highlight the need for methods that do.
Lightning Talk: Luis Mendoza (ISB)
How Well Did You Capture that Ion? Find Out with PeptidePrisoner!
Luis Mendoza, Michael Hoopmann, Eric W. Deutsch, and Robert L. Moritz


In tandem mass spectrometry (MS/MS), peptides are selected for fragmentation from the mixture of ions generated in the first stage of MS. While modern mass spectrometers typically report precursor m/z values with an accuracy of a few ppm, isolation windows are typically much wider and may allow other ions to be co-fragmented. In such a case, the resulting spectra can become complex and difficult to interpret, which can result in decreased specificity and accuracy.
Here we present PeptidePrisoner, a new tool in the Trans-Proteomic Pipeline (TPP) that provides users with an isolation quality score that makes use of the theoretical isotopic envelope of the precursor ion, as well as the total signal attributed to that ion in the selection window.
Lightning Talk: Justin Sanders (UW)
Learning a score function for shotgun proteomics database search
Justin Sanders (1), Sewoong Oh (1), William Stafford Noble (1)(2)

(1) Department of Genome Sciences, University of Washington (2) Paul G. Allen School of Computer Science and Engineering, University of Washington

Computational algorithms for sequence database search are a cornerstone of modern shotgun proteomics experiments, enabling the identification and quantitation of peptides from tandem mass spectra. All existing mass spectrometry database search engines for proteomics use hand-designed score functions to assess the quality of a candidate peptide-spectrum match (PSM). Here, we propose to instead learn this score function directly from data. We train a deep neural network to solve the database search problem by using repository-scale, unlabeled mass spectra and a semi-supervised loss function based on the theory of target/decoy competition. We test the hypothesis that such a model, applied to spectra and peptides not used during training, will yield better statistical power than existing, hand-designed database search score functions. Here we present preliminary results on the feasibility and efficacy of our semi-supervised training objective. While our learned score function assigns peptides to more spectra with high confidence than existing methods, orthogonal validation reveals many of these additional PSMs to be implausible. We investigate a set of potential pitfalls that must be considered when training models using target/decoy analysis which may underlie the observed bias in predictions, and explore strategies to avoid or mitigate them.
Lightning Talk: David Shteynberg (ISB)
PeptideProphet VMC Model for Validation of Rare Peptide PTMs in Complex Samples
David D Shteynberg1; Alex Zelter2; Nina Isoherranen2; Michael R Hoopmann1; Luis Mendoza1; Jimmy Eng2; Eric W. Deutsch1; Robert L. Moritz1

1Institute for Systems Biology, Seattle, WA; 2University of Washington, Seattle, WA

Introduction The Trans-Proteomic Pipeline (TPP) continues to be the gold-standard, open-source analysis suite for proteomics data, and its major component PeptideProphet was first published 21 years ago.
We describe a novel model in PeptideProphet for validating PSM data searched with many variable modifications, some of them rare and others less rare, in one search.
The Variable Modification Count (VMC) model assists PeptideProphet’s classification of PTM containing PSMs. Intuitively, such PSMs are more likely to occur among random results than among correct results.
VMC allows PeptideProphet to compute more accurate FDRs compared to the previous iterations of this model, as indicated by entrapment decoys employed during the search. This is done, while preserving rare PTM containing PSMs with strong spectral evidence.
Methods To test the feasibility of detecting rare PTMs in complex data, we performed native bioactivation of raloxifene in insect cell microsomes, generating a highly complex sample made by co-expressing native human CPY3A4, cytochrome P450 reductase and cytochrome b5 in insect cells.
We collected DDA and DIA spectral data and converted these to mzML.
The mzML data was searched with comet and the resulting pepXML file analyzed with the TPP using PeptideProphet, iProphet and PTMProphet to validate the search results at the PSM level, peptide level and PTM level, respectively.
DIA data was first processed with the TPP tool, DISCo (Data-Independent-Signal-Correlator), that extracts searchable fragment data from DIA data files and generates mzML files that can be searched by comet directly.
Preliminary Data Raloxifene metabolism produces several electrophilic species that form protein adducts of mass 471.1504 Da. We incubated raloxifene with insect cell microsomes resulting in raloxifene metabolism and protein adduct formation. We collected data on unexposed (solvent only), light (d0) raloxifene exposed, heavy (d4) raloxifene exposed and a mixture of d0/d4 raloxifene exposed samples.
Comet searches were performed on DDA data allowing for variable modifications of 471.1504 (d0 raloxifene diquinone methide metabolite) and 475.1755 (d4 raloxifene diquinone methide metabolite) on cysteine, tryptophan and tyrosine, 15.9949 on methionine (oxidation) and 57.021464 on cysteines (carbamidomethyl).
The database used for searching was composed of UniProt protein sequences of the organism Spodoptera frugiperda, plus the human P450 enzymes CYP3A4 and CYP3A5, human P450 reductase, cytochrome b5, yeast enolase, and common protein contaminants.
Two sets of independently randomized decoy sequences were appended to the target database.
Decoy sequences were generated using DeBruijn repeat-preserving randomization, using software within the TPP.
The decoy sequences were randomly interleaved in the FASTA database and used for the comet search. Search results were classified using both Percolator 3.06 and the TPP at the PSM, peptide, and modifications levels, both with and without the use of the novel modification-specific Variable Modification Count (VMC) model.
Novel Aspect Newly added modification specific VMC Model in PeptideProphet improves classification and reduces false positives rates among rare PTM containing peptides as confirmed by both independent decoys and prior knowledge of d0/d4 labeled sample types.
Lightning Talk: Bo Wen (UW)
Learning to predict spectrum quality
Bo Wen, William Stafford Noble

Department of Genome Sciences, University of Washington, Seattle, WA 98195, USA

A longstanding goal in the analysis of tandem mass spectrometry data is to automatically assign a quality score to each observed MS/MS spectrum. Such a score would be useful not only for evaluating the quality of the data but also in the context of assigning peptides to spectra. However, spectrum quality prediction is still challenging due to the complexity of the spectra. In this study, we have developed a method to train a spectrum quality classier using PSMs from the MassIVE-KB dataset by leveraging a spectrum representation learned from deep learning. We demonstrate the performance of this method on a multi-species dataset.
Full Catered Lunch
Session 3: Protein Interactions and Structures
Chair: Devin Schweppe (UW)
Andrew Emili (OHSU)
The Emili Lab for Network Systems Biology: Mapping Macromolecular Interactions for Drug Discovery
Andrew Emili

Knight - OHSU

Will describe new (and ongoing) efforts at the (new) Emili Lab for Network Systems Biology at OHSU aimed at Mapping Macromolecular Interactions for Drug Discovery
Anna Bakhtina (UW)
Quantitative cross-linking mass spectrometry elucidation of membrane protein unfolding to visualize membrane complexes dynamics.
Anna Bakhtina, Sung-Gun Park, Martin Mathay, James Bruce

University of Washington Department of Genome Sciences

Membrane protein complexes perform many of the critical functional processes needed to support life.
However, the study of membrane complex interactions, conformations to decipher how these confer function remains challenging due to the effect of the membrane lipid environment and its importance for maintaining membrane protein conformations and interactions.
Quantitative cross-linking mass spectrometry (qXL-MS) offers unique potential for study of membrane protein interactions, where protein and protein complex architectural features can be captured by covalently linking proximal residues as they exist within native membrane environments.
Here we combine robust quantitative cross-linking with detergent driven protein unfolding to develop a method to study protein complexes. We applied this method to elucidate remodeling of membrane protein complexes during mPTP opening and aging.
Mitochondria were isolated from mouse kidney or liver by differential centrifugation. The mitochondria were split into 8 pools, and 4 pools were treated with increasing concentration of digitonin, while the other fur were used as controls. iqPIR cross-linkers were then added in pairwise manner to controls and digitonin treated samples; reaction was allowed to proceed for 30 min. Mitochondira was then lysed, , digested with trypsin and cross-linked peptides were enriched. Cross-linked peptides were analyzed by LC-MS/MS and quantified based on iqPIR isotope modified peptide and fragment ions.
Log ratios for each cross-link (1, 2, 5, or 10 g of digitonin treatment compared to no digitonin control) were calculated. Mitochondria from young and old mouse tissue, and with or without mPTP inhibitors and activators were processed similarly.
Digitonin is commonly used to extract mitochondrial membrane complexes, including supercomplexes. Our preliminary data show that cross-linked peptides that identify supercomplex assemblies (for example CI-CIV cross-linked species) are altered with increased concentration of digitonin.
qXL-MS enabled the visualization of supercomplex assembly by recording changes in quantitation of the cross-links with increased detergent concentration.
Supercomplex formation is expected to be altered with age-related mitochondrial functional decline.
A comparison of cross-link level changes during digitonin induced unfolding showed a significant difference in supercomplex stability in young and aged mouse mitochondria.
We also applied Triton-X, another non-ionic detergent used for membrane protein extraction, to isolated mitochondrial membranes in the presence of either an inhibitor or activator of mitochondrial permeability transition pore (mPTP) opening. The mPTP is a large multiprotein non-specific channel that opens during mitochondrial stress and allows for loss of important solutes and dissipation of mitochondrial membrane potential.
The composition and mechanism of mPTP opening has long been studied with traditional methods, but still remain controversial.
Treatment of isolated mitochondria with calcium ions induces structural rearrangement and opening of the mPTP.
This presentation will highlight cross-link level changes recorded with and without calcium.
To further investigate putative mPTP associated cross-links mitochondria were pretreated with ADP, mPTP opening inhibitor, and then treated with calcium. These results illustrate how qXL-MS can provide unique insight on membrane protein conformations and interactions not possible with other techniques. In conclusion, we have developed a novel method to study membrane protein and protein complexes dynamics combining qXL-MS and detergent induced disruption of protein-protein interactions.
Lindsey Ulmer (UW)
Precision Crosslinking: Identifying and Interpreting Crosslinks Originating from Photoactive Amino Acids in Less-Ordered Regions of Proteins
Lindsey D. Ulmer,1 Christopher N. Woods,2 Natalie L. Stone,2 Rachel E. Klevit,2 Matthew F. Bush1

1 Department of Chemistry, University of Washington, Seattle, WA 98195, United States 2 Department of Biochemistry, University of Washington, Seattle, WA 98195, United States

Less-ordered elements of proteins are often absent from structures determined using condensed-phase techniques, including X-ray crystallography and cryo-EM. Benzoylphenylalanine (BPA) can be incorporated as a noncanonical amino acid using molecular biology and photoactivated to form crosslinks with any amino acid. Here, we use BPA to probe a region of the small heat shock protein HSPB5 that has resisted molecular-level structural characterization. However, the ability of BPA to react with any amino acid also makes it more challenging to identify and interpret crosslinks. We developed a workflow using the Trans-Proteomic Pipeline (TPP) to identify high-confidence, residue-level crosslinks. We also established a statistical strategy using bootstrapping to aid in the interpretation and validation of crosslinking patterns. The chemistry of BPA – photoactivated BPA can form covalent bonds to all amino acids – presents several challenges to standard crosslinking workflows. We developed an interactive notebook that combines results from different TPP tools to identify high-confidence, residue-level BPA crosslinks. We characterized the performance of this identification workflow by varying the size of the protein database used for crosslink searches. This workflow outperforms StavroX, a program that has previously been used to identify BPA crosslinks, by identifying far more crosslinks in far less time at a given false-discovery rate. We applied this identification workflow to data for six mutants of HSPB5 that each incorporate BPA at a different site.
We then used our bootstrapping method to evaluate the similarity between the crosslinking patterns originating from those six sites. In bootstrapping, a probability distribution function (PDF) is resampled with replacement many times, a test statistic is calculated for each resample, and the distribution of the resulting test statistics is used to establish thresholds for statistical significance. Using experimental datasets as the PDF and the Jensen-Shannon similarity score as the test statistic, we established that the crosslinks from different BPA sites are significantly different than each other (p ≤ 5⸱10–4).
We also applied our bootstrapping strategy to evaluate whether the observed crosslinks could be explained by factors other than structure, e.g., crosslinking dominated by reactivity (represented by a PDF that depends on the intrinsic reactivity of each type of amino acid) or random interactions (represented by a uniform PDF). Using variance and maximum density as the test statistics, we established that the observed crosslinks are significantly different than would be expected from either alternative PDF (p << 10–4). This result supports our hypothesis that these experiments are very sensitive to the structure of HSPB5, even in “disordered” regions that have proven to be especially resistant to structural characterization. We propose that this bootstrapping strategy will also be useful for interpreting the results from other solution-labelling experiments.
Avik Basu (OHSU)
Proximity labeling to study key protein-protein interactions at synapse
Avik Basu1*, Yuan Tian2*, Sadhna Phanse1, Heng-Ye Man2, and Andrew Emili1,2,

1. Department of Chemical Physiology and Biochemistry, Division of Oncological Sciences, Oregon Health and Science University, Portland, OR, USA 2. Department of Biology, Boston University, Boston, MA, USA

To investigate the molecular basis of homeostatic synaptic plasticity, we adapted a photo-proximity labeling-based functional proteomics workflow to identify protein-protein interactions involving the GluA1 subunit of AMPA receptor (AMPAR) in live primary rat neurons. Using antibodies conjugated to a photoactivatable flavin-based catalyst, we demonstrated target selective biotinylation and recovery of AMPAR along with both well described and previously unreported auxiliary proteins associated with neurotransmission. This resulted in the identification and functional characterization of the calcium sensor NCS1, which we validated as a novel regulator of homeostatic plasticity initiated via the calcium-permeable form of AMPAR.
Lightning Talk: May Constabel (UW)
Programmable, Temperature-Controlled ESI Enables Online Thermal Cycling and Disulfide Bond Reduction of Proteins
May A. Constabel, Christopher J. Weir, Theresa A. Gozzo, Meagan M. Gadzuk-Shea, and Matthew F. Bush

Department of Chemistry, University of Washington

Introduction We developed a low-thermal-mass, programmable, temperature-controlled electrospray ionization (ptESI) source capable of rapid thermal modulation of ESI samples. With ptESI, we can increase or decrease the temperature of the solution within the nanospray capillary during experiments at rates of at least 0.5 °C·s–1 with high fidelity. Combined with ion mobility (IM) mass spectrometry (MS), ptESI enables us to probe protein folding/unfolding, conformational stability, and protein aggregation. Here, we demonstrate two applications of ptESI: thermal cycling to characterize the reversibility of protein folding and online disulfide bond reduction of proteins.
Methods Protein Unfolding with Rapid Cycling: Solutions of 10 µM myoglobin were prepared in 200 mM ammonium acetate. 4-5 µL of solution was loaded into nESI capillaries (borosilicate glass pulled to 1-3 um i.d.), a capillary inserted into the small copper block of the ptESI source and sprayed. The temperature was cycled 4 times continuously between 30 to 90 °C using a sawtooth function with slopes of ±0.5 °C·s–1 while MS spectra were continuously acquired using a Waters Cyclic IM-MS system. Online Disulfide Bond Reduction: Samples containing ribonuclease A (RBA) or cytochrome c were exchanged into aqueous ammonium acetate and then diluted with a solution containing a reducing agent to a final concentration of 10 µM RBA and varied concentrations of DTT, βME, or TCEP. The protein solutions were cycled once from 30 to 80 °C at a rate of 0.5 °C·s–1. Additional RBA solutions were incubated at 37 °C for 1 hour or at room temperature for 1 hour prior to analysis.
Preliminary Data Protein Unfolding with Rapid Cycling: Myoglobin (Mb) is a heme-binding protein that has previously been reported to have reversible unfolding with heating. For samples at 30 °C, the mass spectrum is dominated by peaks assigned to +8 and +9 holo-Mb. With increasing temperature, the intensity of those peaks decreases and those for +12 to +18 apo-Mb increase. When the temperature of the sample is cooled, the original peaks are once again the most intense. That process corresponds to the first cycle. With increasing cycle number, the spectra at low temperatures exhibit increasing relative intensity for both +8 to +14 apo-Mb ions and heme (616 m/z). Therefore, the unfolding of Mb can be associated with reversible and irreversible changes in structure. Incorporation of a complementary IM dimension is underway.
Online Disulfide Bond Reduction: The addition of 5-10 mM DTT or 10 mM βME was sufficient to irreversibly reduce all four native disulfide bonds in RBA during the first 50 s of one thermal cycle. Mass spectra were continuously acquired during the cycle; a clear shift to higher m/z corresponding to the addition of 8 hydrogens was observed. The capability to perform in-source reduction can reduce experiment time by eliminating the long incubation sample preparation step. As a control, reducing agents were also added to solutions of cytochrome c, a protein with no native disulfide bonds. No changes in the neutral mass of cytochrome c were observed. Nonspecific adducts of DTT and βME were not observed; however, adducts of TCEP were observed. This is being investigated further.
Lightning Talk: Alice Martynova (UW)
Solving the Ligand Puzzle: Native Mass Spectrometry and Counting Ligands Bound to Fatty Acid Binding Proteins
Alice Martynova, King C. B. Yabut, Benjamin P. Zercher, Nina Isoherranen, and Matthew F. Bush

University of Washington, Department of Chemistry University of Washington, School of Pharmacy, Department of Pharmaceutics

Native mass spectrometry (MS) has emerged as a powerful technique for analysis of intact proteins and protein-ligand complexes in solution. Native MS can provide unique insight into binding interactions, stoichiometry, interferences, and structural dynamics, which can all contribute to a more complete understanding of complex binding processes. In this study, we employed native MS for the analysis of fatty acids binding proteins (FABPs) and their protein-ligand complexes. FABPs are a family of small cytosolic proteins ubiquitously expressed in tissues that play a crucial role in the transport and metabolism of fatty acids and lipid-like molecules. In addition to their roles in lipid metabolism, FABPs have been implicated in various physiological processes, such as inflammation, insulin resistance, and cancer, which makes them intriguing targets for therapeutic and diagnostic purposes. Despite a well-established role in lipid metabolism, the involvement of FABPs in drug metabolism remains undercharacterized. Using native MS, we aimed to shed light on the drug binding capacity and drug metabolism-related functions of FABP1, specifically known as liver-type FABP.
Initial characterization of intact, recombinant FABP1 revealed that established methods for expressing, purifying, and delipidating this protein results in the retention of significant populations of copurifying molecules, presumably fatty acids from the expression system (E. coli). This finding has significant implications for interpretation of previous FAPB1 ligand binding assays, which may have probed competition between target ligands and copurifying molecules in additional to the binding of the target ligand to apo FABP1. To address this, we employed native MS to evaluate the effectiveness of different delipidation methods. Additionally, we observed evidence for a 60 Da modification to FABP1, which we tentatively assigned to modification of the single cysteine in the binding pocket of FABP1. We investigated the effectiveness of the reducing agent dithiothreitol (DTT) on this modification and its impact on native MS characterization.
Subsequently, we analyzed FABP-ligand complexes using diclofenac and 11-dansylaminoundecanoic acid (DAUDA). The analysis elucidated the stoichiometry and provided direct evidence for both single and double addition of DAUDA to FABP1, which has a single binding cavity. The characterization of the interactions between FABP1 and diclofenac, a widely used nonsteroidal anti-inflammatory drug (NSAID), expands our understanding of the binding interactions as a result drug metabolism. Additionally, we validated the use of DAUDA, a fluorescent probe, for fluorescence displacement assays, which proved useful for studying protein-drug complexes. Subsequently, the fluorescent displacement assay was applied to investigate various protein-drug complexes.
These findings provide valuable insights into the drug binding properties of FABP1 and its potential involvement in drug metabolism, opening avenues for further research and the development of therapeutic strategies targeting FABPs. More generally, these experiments highlight the benefits using native MS for characterizing protein-ligand interactions, including facile observation of interfering molecules and differentiation between proteins with one or two ligands bound.
Keywords: Fatty acid binding proteins, native mass spectrometry, drug metabolism, protein-drug interactions, fluorescent displacement assay.
Session 4: Proteomes, Metaproteomics & Metabolomics
Chair: Brook Nunn (UW)
Miranda Mudge (UW)
Predicting harmful algal blooms through high-resolution temporal metaproteomics of a bacterial microbiome
Miranda Mudge, Alan Min, Emma Timmins-Schiffman, Gabriella Chebli, Julia Kubanek, Michael Riffle, William Noble, Brook Nunn

Molecular and Cellular Biology Department, University of Washington; Department of Genome Sciences, University of Washington; School of Biological Sciences, Georgia Institute of Technology

The San Juan Islands, Washington, USA border a dynamic coastal ecosystem that is unique to the world in its predictable occurrence of biannual harmful algal blooms (HABs). To fully characterize this ecosystem and harness its potential for revealing clues to HAB initiation, a high-resolution temporal study was conducted to identify microbial community interactions and their molecular-level controls on large scale algal bloom events. In June of 2021 we sampled coastal waters in Eastsound by collecting the bacterial microbiome every 4 hours for 22 days, with additional time-matched samples for environmental and water chemistry analyses. The metagenome was sequenced and a quantitative metaproteomic analysis performed on the last 6 days of the timeseries during a period characterized by low environmental disturbance preceding a Chaetoceros bloom. We analyzed the microbiome metaproteome on a Lumos mass spectrometer using Data Independent Acquisition (DIA) and processed samples using Prosit, EncyclopeDIA, and Skyline. Peptide identifications were inferred using an innovative pipeline in which samples initially processed using Data Dependent Acquisition (DDA) were searched against a time- and location-specific metagenome to create a sample-specific DDA-filtered metagenomic library calibrated for searching the DIA samples with high confidence. The microbiome revealed a rapid shift in profile prior to the initiation of the HAB, characterized by an increase in peptides attributed to the bacterial class Verrucomicrobiae, previously documented for their potential to degrade complex HAB biproducts. Our time series of a marine bacterial community is unprecedented in its scope and resolution and has revealed that some bacterial taxa are metabolically linked to the formation of a HAB, yielding potentially specific and predictive biomarkers of bloom formation.
Aivett Bilbao (PNNL)
Using PeakDecoder to decipher microbial omics signatures in multidimensional mass spectrometry
Aivett Bilbao, Nathalie Munoz, Yuqian Gao, Joonhoon Kim, Marija Velickovic, Chaevien S. Clendinen, Daniel J. Orton, Kunal Poorey, Kyle R. Pomraning, Karl Weitz, Meagan Burnet, Carrie D. Nicora, Rosemarie Wilton, Shuang Deng, Ziyu Dai, Ethan Oksen, Aaron Gee, Rick A. Fasani, Anya Tsalenko, Deepti Tanjore, James Gardner, Richard D. Smith, Joshua K. Michener, John M. Gladden, Erin S. Baker, Christopher J. Petzold, Young-Mo Kim, Alex Apffel, Jon K. Magnuson, and Kristin E. Burnum-Johnson

Pacific Northwest National Laboratory, Richland, WA, USA; US Department of Energy, Agile BioFoundry, Emeryville, CA, USA; Sandia National Laboratory, Livermore, CA, USA; Argonne National Laboratory, Lemont, IL, USA; Lawrence Berkeley National Laboratory, Berkeley, CA, USA; Agilent Research Laboratories, Agilent Technologies, Santa Clara, CA, USA; Oak Ridge National Laboratory, Oak Ridge, TN, USA; Department of Chemistry, University of North Carolina, Chapel Hill, NC, USA

Multidimensional mass spectrometry (MS) data, including liquid chromatography (LC) and ion mobility (IM) separations, and collected in data-independent acquisition (DIA) mode, contains rich information to study molecular networks of cellular processes and biochemical reactions in living systems. However, this kind of data requires more sophisticated processing algorithms compared to traditional MS. We recently demonstrated PeakDecoder, a new algorithm that uses a support vector machine model to enable metabolite annotation and accurate profiling in multidimensional MS data. We studied 64 metabolites from several fungal and bacteria strains relevant in the biotechnology field for production of value-added chemicals, including biofuels. PeakDecoder enabled the interpretation of 2,683 metabolite features across 116 microbial samples. This presentation will describe the algorithm and performance of PeakDecoder, and the progress towards PeakDecoder2 adapting multimodal data fusion deep learning models with convolutional neural networks for metabolomics and proteomics. For the analytical workflow, LC methods were optimized (C18 column with 30 min gradient for proteomics and HILIC column with 7 min gradient for metabolomics) and coupled to an Agilent 6560 Drift Tube Ion IM-QTOF-MS operated in All ions DIA mode. Data was acquired in positive ionization with ramped collision energy values (10–40V) for proteomics and negative ionization with 20V or 40V for metabolomics. Finally, results using public experimental and predicted libraries to process data from a research project aiming to understand plant matter degradation in environmental microbial systems will be presented.
Lightning Talk: Wentao Zhu (UW)
Combined ingestion of polystyrene microplastics and epoxiconazole increases health risk to mice: Based on their synergistic bioaccumulation in vivo
Wentao Zhu, Daniel Raftery

Northwest Metabolomics Research Center, University of Washington, Seattle, WA 98109, United States of America

Microplastic and pesticide are two common environmental pollutants whose adverse effects have been widely reported, but it is unclear whether they cause combined toxicity in mammals. In this study, polystyrene microplastics (5 µm, 0.012 or 0.120 mg/kg) or/and epoxiconazole (0.080 mg/kg) were administered orally to mice for 6 weeks, their toxicity to liver and kidney was assessed from changes in histopathology, tissue function, oxidative defense system and metabolic profile. In addition, mechanism of combined toxicity was explored in terms of bioaccumulation levels, intestinal barrier, gut microbiota. Results showed that combined ingestion of polystyrene (0.120 mg/kg) and epoxiconazole caused more severe tissue damage, dysfunction, oxidative stress, and metabolic disorders compared to single exposure sources. Interestingly, occurrence of combined toxicity was associated with their increased accumulation in tissues. In-depth exploration found that epoxiconazole caused intestinal barrier damage by targeting the gut microbiota, leading to massive invasion and accumulation of polystyrene, which in turn interfered with the metabolic clearance of epoxiconazole in liver. In all, findings highlighted that polystyrene and epoxiconazole could cause combined toxicity in mice through the synergistic effect of their bioaccumulation in vivo, which provided new reference for understanding the health risks of microplastics and pesticides and sheds light on the potential risk to humans of their combined ingestion.
Lightning Talk: Tami Leppert (ISB)
Mapping the Arabidopsis thaliana proteome in PeptideAtlas and the nature of the unobserved (dark) proteome; strategies towards a complete proteome.
Klaas J. van Wijk^, Tami Leppert+, Zhi Sun+, Alyssa Kearly*, Margaret Li+, Luis Mendoza+, Isabell Guzchenko^, Erica Debley^, Georgia Sauermann^, Pratyush Routray^, Sagunya Malhotra+, Andrew Nelson*, Qi Sun% and Eric W. Deutsch+

^Section of Plant Biology, School of Integrative Plant Sciences (SIPS), Cornell University, Ithaca, NY 14853, USA; +Institute for Systems Biology (ISB), Seattle, Washington 98109, USA; *Boyce Thompson Institute, Ithaca, NY 14853.; %Computational Biology Service Unit, Cornell University, Ithaca, NY 14853.

We present a new release of the Arabidopsis thaliana PeptideAtlas proteomics resource providing protein sequence coverage, matched mass spectrometry (MS) spectra, selected PTMs, and metadata. 70 million MS/MS spectra were matched to the Araport11 annotation, identifying ∼0.6 million unique peptides and 18267 proteins at the highest confidence level and 3396 lower confidence proteins, together representing 78.6% of the predicted proteome. Additional identified proteins not predicted in Araport11 should be considered for building the next Arabidopsis genome annotation. This release identified 5198 phosphorylated proteins, 668 ubiquitinated proteins, 3050 N-terminally acetylated proteins and 864 lysine-acetylated proteins and mapped their PTM sites. MS support was lacking for 21.4% (5896 proteins) of the predicted Araport11 proteome – the ‘dark’ proteome. This dark proteome is highly enriched for certain (e.g. CLE, CEP, IDA, PSY) but not other (e.g. THIONIN, CAP,) signaling peptides families, E3 ligases, TFs, and other proteins with unfavorable physicochemical properties. A machine learning model trained on RNA expression data and protein properties predicts the probability for proteins to be detected. The model aids in discovery of proteins with short-half life (e.g. SIG1,3 and ERF-VII TFs) and completing the proteome. PeptideAtlas is linked to TAIR, JBrowse, PPDB, SUBA, UniProtKB and Plant PTM Viewer.
Lightning Talk: Gil Omenn (Umich)
The 2023 Report on the Human Proteome from the HUPO Proteome Project
Gilbert S Omenn1, 2, Lydie Lane3, Christopher M Overall4, Cecilia Lindskog5, Charles Pineau6, Nicolle H Packer7, Susan Weintraub8, Sandra Orchard9, Michael Roehrl10, Ed Nice11, Tiannan Guo12, Jennifer E Van Eyk13, Siqi Liu14, Nuno Bandeira15, Ruedi Aebersold16, Robert L. Moritz2, and Eric W Deutsch2

1University of Michigan, 2Institute for Systems Biology, 3CALIPHO Group, SIB Swiss Institute of Bioinformatics, 4University of British Columbia, 5Uppsala Universitet, 6University Rennes, Biosit, 7Macquarie University, 8University of Texas Health Science Center, 9EMBL-EBI, 10Memorial Sloan Kettering Cancer Center, 11Monash University, 12Westlake University, 13Cedars Sinai, 14BGI Group, 15University of California-San Diego, 16ETH-Zurich

Background Since 2010, the Human Proteome Project (HPP) home of the flagship initiative of global HUPO has pursued two goals: (1) To credibly identify the protein parts list and (2) To make proteomics an integral part of multi-omics studies of human health and disease methods.
Methods International collaboration, data sharing, standardized reanalysis of MS data sets by PeptideAtlas and MassIVE with HPP Guidelines for quality and assurance, plus extensive use of antibody profiling by the Human Protein Atlas.
Results According to the neXtProt release of 2023-03, protein expression has now been credibly detected (PE1) for 18,397 of the 19,778 predicted proteins coded in the human genome (93%). Of these PE1 proteins, 17,453 were detected with mass spectrometry in accordance with HPP Guidelines and 944 by a variety of non-MS methods. Conversely, the number of neXtProt PE2, PE3, and PE4 missing proteins has been reduced to 1381. These numbers represent experimental progress on the Human Proteome parts list across all of the chromosomes with significant re-classifications. Meanwhile, there are several categories of predicted proteins that have proven resistant to detection. Applying proteomics with a large array of biological and clinical studies ensures integration with other omics platforms as reported by the Biology and Disease-driven HPP teams.
The HPP has now launched its Grand Challenge to find the functions of every protein.
Conclusions The global proteomics community has made remarkable progress in detecting and characterizing protein expression and protein functions in pathways and networks critical to understanding human health and disease.
Beer and Wine and Tapas Reception with Posters
Thermo Fisher Scientific Hospitality Night: Seattle Mariners vs. Minnesota Twins at T-Mobile Park