Posters - Schedules
Poster presentations at ISMB/ECCB 2021 will be presented virtually. Authors will pre-record their poster talk (5-7
minutes) and will upload it to the virtual conference platform site along with a PDF of their poster beginning July 19
and no later than July 23. All registered conference participants will have access to the poster and presentation
through the conference and content until October 31, 2021. There are Q&A opportunities through a chat
function and poster presenters can schedule small group discussions with up to 15 delegates during the conference.
Information on preparing your poster and poster talk are available at:
https://www.iscb.org/ismbeccb2021-general/presenterinfo#posters
Ideally authors should be available for interactive chat during the times noted below:
View Posters By Category
Session A: Sunday, July 25 between 15:20 - 16:20 UTC |
Session B: Monday, July 26 between 15:20 - 16:20 UTC |
|---|---|
Session C: Tuesday, July 27 between 15:20 - 16:20 UTC |
Session D: Wednesday, July 28 between 15:20 - 16:20 UTC |
|---|---|
Session E: Thursday, July 29 between 15:20 - 16:20 UTC |
|---|
Short Abstract: Recent enhancements in the sequence speed and sensitivity of mass spectrometry (MS)-based proteomics by the introduction of the Parallel Accumulation–Serial Fragmentation (PASEF) method [Meier, F. et al., 2018] have drawn more attention to the timsTOF Pro device (Bruker Daltonik) [Beck, S. et al., 2015]. However, accession, processing and visualization of the high-throughput five-dimensional LC-TIMS-QTOF data remains challenging. To simplify these time-consuming steps we have developed several open-source Python packages with stand-alone graphical user interfaces (GUI) that enable users to visualize experimental data quickly, easily and interactively without requiring programming skills.
Here we present three of these tools: AlphaTims, AlphaViz and AlphaMap. AlphaTims provides an easily installable GUI to visualize and explore unprocessed Bruker data in milliseconds. Once this data is processed with e.g. AlphaPept, MaxQuant [Cox, J., Mann, M., 2008], Spectronaut or DIA-NN [Demichev, V. et al., 2020], AlphaViz provides an automated visualization pipeline to look at the analyzed data and easily assess the quality of the overall samples or to evaluate the quality of raw data for proteins of interest. Finally, AlphaMap enables visual exploration of MS data on a peptide level with UniProt resources, including currently known PTMs, proteins domains, etc.
Short Abstract: Crosslinking is used to study the structure of single-proteins, multi-protein complexes, or protein-protein interactions. The identification of cross-linked peptides by conventional database search has quadratic complexity as any pair of peptides needs to be considered. Yet, the detection of a peptide being cross-linked based on its tandem mass spectrum and prior to the search has the potential to speed up the identification process because only promising spectra need to be searched. Here, we detect if peptides are cross-linked (in contrast to being linear peptides) based on their tandem mass spectrum using our deep learning approach, called AHLF (“ad hoc learning of peptide fragmentation”). AHLF is able to learn less studied patterns in spectra by not relying on explicit domain knowledge that may bias or limit the characterization of tandem mass spectra. To elevate unrestricted learning from spectra, our deep learning model was end-to-end trained on 19.2 million spectra. The original task of AHLF was phosphopeptide detection, but here we demonstrate a transfer learning approach for AHLF. Cross-linked peptides are detected with an area under the receiver operating characteristic curve (AUC) of up to 94%. AHLF is flexible and allows spectrum-based decision making which can be integrated to existing proteomics pipelines.
Short Abstract: Retention time (RT) alignment is one of crucial steps in quantitative proteomics workflows. Currently, many alignment methods use a certain linear or nonlinear fitting function to align RT, but few of they perform well on complex data sets. Some methods further incorporate MS/MS information to improve accuracy. But they have a limitation to low abundance peptides which usually get low quality MS/MS spectra. Here, we present a deep learning based RT alignment model, named DeepRTAlign independent of MS/MS information. After a preliminary alignment step by calculating the average time shift between two samples, we use RT, m/z, charge, fraction, intensity as features to train our deep learning model. Our model shows a better performance on a complex data set (early stage hepatocellular carcinoma, AUC: 0.99, precision: 0.95, recall: 0.96) compare with one of the best RT alignment tool MZmine 2 (precision: 0.92, recall: 0.92). As a deep learning based tool, DeepRTAlign does not need users to predefine any parameter, and can easily be adapted to other similar data such as those from gas chromatography. Moreover, our model does not need MS/MS information, which may be useful in finding low abundance candidate biomarkers.
Short Abstract: Enzymes dictating the biophysical outcomes of cells are controlled through allosteric interaction with metabolites. While various protein-metabolite allosteric interactions have been identified and described, doing so en masse is difficult and time-consuming. We previously developed a high-throughput method for the identification of these interactions, “mass spectrometry integrated with equilibrium dialysis for the discovery of allostery systematically”, or “MIDAS”. We present Electrum, a web app for analyzing this unique data type. Electrum provides a standard template for this type of data and includes various features that enable the exploration and analysis of these data. To aid in the biological contextualization of these data, we include the ability to query these protein-metabolite interactions for relevant information from databases such as the Human Metabolome Database and the Reactome Knowledgebase. To facilitate the identification of regulatory-relevant patterns, we developed tools that systematically identify statistical sub-structure enrichments within all the metabolites interacting with a target protein. Conserved substructures can indicate the nature and mechanism of these allosteric interactions. To our knowledge, Electrum is the first tool to provide such a utility. The Electrum web portal can be accessed at rutter.chpc.utah.edu/Electrum and the source code for Electrum is available at github.com/Electrum/Electrum-app under a GPL-3.0 license.
Short Abstract: For the identification of peptides in a sample by tandem mass spectra, as generated by data-dependent acquisition, protein sequence databases like the well known UniProtKB, provide the basis for most spectrum identification search engines. In addition, for targeted proteomics approaches like selected reaction monitoring and parallel reaction monitoring, knowledge of the peptide sequences, their masses, and whether they are unique for one protein in a database is essential. Because most bottom-up proteomics approaches use trypsin to cleave the proteins in a sample, the tryptic peptides contained in a protein database are of great interest. We recently published MaCPepDB (mass-centric peptide database), that consists of the complete tryptic digest of the Swiss-Prot and TrEMBL parts of UniProtKB. This database is especially designed to query peptides by sequence or mass with additional filters, like mass tolerance or posttranslational modifications, and return the respective annotated peptides.
With increasing number of queries and peptides, MaCPepDB quickly reached the limits of its original design.
The enhanced version of MaCPepDB takes advantage of the scalable database Citus. With the help of Citus, MaCPepDB is able to support a higher amount of concurrent queries and a general speedup of queries by distributing the load on multiple servers.
Short Abstract: Top-down proteomics (TDP) is gaining great interest in biological, clinical, and medical sciences, as the method of the choice to study proteoforms. While significant improvements have been made on different aspects of TDP protocols, data-dependent acquisition (DDA) has been optimized for bottom-up proteomics, not for TDP. Dedicated acquisition methods thus have the potential to greatly improve TDP.
We present FLASHIda, an intelligent data acquisition method for TDP that ensures the selection of high-quality precursors of diverse proteforms. FLASHIda interfaces with Thermo Scientific iAPI that provides MS1 full scans real-time. By transforming the raw m/z-intensity spectrum to mass-quality spectrum instantly with FLASHDeconv and using a machine learning technique assessing the signal quality, FLASHIda implements Top-N high-quality precursor mass acquisition with a quality-based mass exclusion.
In the benchmark tests with E. coli lysate 90-min gradient single runs (nano-RPLC, Thermo Scientific Orbitrap Eclipse), FLASHIda almost doubled the unique proteoform count (~1,600) as compared with the standard acquisition (~820). Alternatively, similar numbers (~800) as with standard DDA were reported in FLASHIda runs on drastically shorter gradient runs (30-min).
Since FLASHIda does not require any modification in experimental set-ups, it could be readily adopted for TDP study of complex samples to raise proteoform identification sensitivity.
Short Abstract: Tandem mass spectrometry (MS/MS) can be used to quantify thousands of peptides in a complex biological mixture. However, these quantitative measurements depend in part on the properties of the peptide sequence. We propose a machine learning approach to quantify and eliminate these peptide-specific artifacts. We model the observed peptide intensity as a composition of a peptide coefficient and an adjusted abundance. We then base our model on the key assumption that sibling peptides (i.e., peptides that co-occur in the same protein) should have equal abundances. Accordingly, we implement a neural network with a Siamese architecture to learn peptide coefficients from amino acid sequences, which is trained to minimize the distance between adjusted abundances of sibling peptides. We demonstrate that peptide coefficients are consistent across different MS/MS runs and that the coefficients inferred from one set of runs can generalize to other runs. We aim to extend our prediction model to new proteins and datasets to eliminate these peptide-specific effects, thereby yielding more accurate quantification.
Short Abstract: The isoelectric point is the pH at which a particular molecule is electrically neutral due to the equilibrium of positive and negative charges. In proteins and peptides, this depends on the dissociation constant (pKa) of charged groups of seven amino acids and NH+ and COO- groups at polypeptide termini. Information regarding isoelectric point and pKa is extensively used in two-dimensional gel electrophoresis (2D-PAGE), capillary isoelectric focusing (cIEF), crystallisation, and mass spectrometry. Therefore, there is a strong need for the in silico prediction of isoelectric point and pKa values. In this work, I present Isoelectric Point Calculator 2.0 (IPC 2.0), a web server for the prediction of isoelectric points and pKa values using a mixture of deep learning and support vector regression models. The prediction accuracy (RMSD) of IPC 2.0 for proteins and peptides outperforms previous algorithms: 0.848 versus 0.868 and 0.222 versus 0.405, respectively. Moreover, the IPC 2.0 prediction of pKa using sequence information alone was better than the prediction from structure-based methods (0.576 vs. 0.826) and a few folds faster. The IPC 2.0 webserver is freely available at www.ipc2-isoelectric-point.org
Short Abstract: Mass spectrometry-based proteomics is a key technique for analyzing proteins in biological samples. With more and more mass spectrometry data becoming easily accessible and the ever-increasing capabilities of machine learning methods, the mass spectrometric life cycle of a protein and its peptides can now be reliably predicted. The integrated application of these new prediction tools is currently hindered by the various specialized software solutions and their different requirements. We present millipede, a versatile deep learning python package to predict relevant protein and peptide properties for simulating the entire mass spectrometric life cycle. The prediction tasks range from protein digestion, charge prediction, and retention time prediction to MS1 and MS2 spectra prediction. Millipede sets itself by being first, an all-in-one software solution that can be used to compute features for dedicated peptide rescoring algorithms, such as percolator and second, a software solution to simulate mass spectrometry acquisitions. Millipede therefore enables high quality simulated mass spectrometric data that will improve benchmarking, testing, and optimization of algorithms and further guide instrument parameter choices.
Short Abstract: Data Independent Acquisition (DIA) is a method of choice for large-scale proteomics experiments. In general, label-free approaches are used to quantify peptides with area under the chromatographic peaks. Hence quantification is reliant on coordinates of the peak in SWATH map, peak-picking software and FDR control strategies. In large studies, changes in LC-MS/MS setup can affect the data-acquisition. In addition, chromatography retention time (RT) coordinate may vary across days. These effects may result in missing values in the final quantification matrix. Thus, in large-scale studies obtaining accurate peptide quantification is challenging.
I have developed a RT alignment tool, DIAlignR, which aligns MS/MS chromatograms using either reference-based or reference-free ways across runs and provides quantification to fill missing values. Since we use raw chromatograms, the alignment is robust compared to existing methods. We tested its performance on 227 runs from multi-laboratory studies [Collins 2017] which included 11 sites across the globe. Each sample had HEK293 cell lysate with iRT peptides and six AQUA peptides for which concentration was known. In total, there were 30 peptides spanning a dynamic range of five orders of magnitude. In this dataset, the alignment has significantly improved quantification for 26 AQUA peptides.
Short Abstract: In sensitive high-throughput LC-MS/MS-based proteomics, information about the presence and stoichiometry of post-translational modifications are normally not readily available. Here we introduce multiFLEX-LF, a computational tool which overcomes this problem. The tool builds upon the previously developed FLEXIQuant and FLEXIQuant-LF workflows, which detect modified peptides and quantify their modification extent by monitoring the differences between observed and expected intensities of the unmodified peptides. To this end, FLEXIQuant-LF and multiFLEX-LF utilize robust linear regression to calculate the degree of differential modification of unmodified peptides relative to a within study reference. multiFLEX-LF was developed to perform FLEXIQuant-LF on entire label-free LC-MS/MS-based discovery proteomics datasets. To detect modification dynamics and co-regulated modifications, the peptides of all proteins are clustered hierarchically based on their computed relative modification scores across the entire dataset.
multiFLEX-LF was applied to a cell-cycle time series proteomics dataset acquired using data-independent acquisition routines. After clustering of the peptides, several clusters of peptides with different modification dynamics were determined.
Overall, multiFLEX-LF enables fast identification of potentially differentially modified peptides and quantification of their differential modification extent in large datasets. multiFLEX-LF can facilitate large-scale research regarding relative changes in the modification extent of peptides in time series and case-control studies.
Short Abstract: Automated peak picking in LC-HRMS-based metabolomics and especially in untargeted approaches is of utmost importance. It is among the first steps of any data processing pipeline and aimed at recognizing true chromatographic peaks. Even within a single LC HRMS chromatogram the many recorded compounds can possess vastly different chromatographic peak shapes (e.g. tailing).
In this respect we present PeakBot, a machine-learning model that recognizes chromatographic peaks of different forms and shapes. It operates on the profile-mode data, thus enabling the analysis of a richer dataset and improving prediction quality. Using a training dataset consisting of some 100 human-curated chromatographic peak profiles as well as non-peak containing regions of the chromatogram, PeakBot augments and extends the dataset by randomly combining them into complex areas and thereby generating a large training-set with the aim of correctly detecting the target peaks. After training with the augmented training dataset, PeakBot can be used to detect other chromatographic peaks. PeakBot’s performance competes with other tools with respect to prediction accuracy and speed. It reports a category, the retention time, m/z value, bounding box as well as a binary mask for detected chromatographic peaks. PeakBot is open-source, implemented in Python and uses TensorFlow.
Short Abstract: Identification of small molecules in a biological sample remains a major bottleneck in molecular biology, despite a decade of rapid development of computational approaches for predicting molecular structures using mass spectrometry (MS) data. Recently, there has been increasing interest in utilizing other information sources, such as liquid chromatography (LC) retention time (RT), to improve identifications solely based on MS information, such as precursor mass-per-charge and tandem mass spectrometry (MS2).
We put forward a probabilistic modelling framework to integrate MS and RT data of multiple features in an LC-MS experiment. We model the MS measurements and all pairwise retention order information as a Markov random field and use efficient approximate inference for scoring and ranking potential molecular structures. Our experiments show improved identification accuracy by combining MS2 data and retention orders using our approach, thereby outperforming state-of-the-art methods. Furthermore, we demonstrate the benefit of our model when only a subset of LC-MS features has MS2 measurements available besides MS1.
Short Abstract: Data-dependent acquisition (DDA) proteomics is a widely used, however a noisy proteome measurement technique. This is a major obstacle for possible future clinical applications. We therefore aim at learning to denoise and impute data-dependent acquisition proteomics data, using an assembled dataset of replicates of HeLa measurements and proteomics tissue measurements. An unsupervised deep learning approach is used, which aims at finding a common, latent representation of the data which contains all relevant information for denoising the data and imputing missing protein intensities.
Preliminary results will be presented: First, it is discussed which level of the proteomics data is best suited to perform a final protein imputation task. Second, it is considered if biological differences are attenuated by the imputation procedure. Third, the performance transferring models from one tissue cell line to different tissues is assessed. The potential alternative is training a model for each new tissue dataset. Fourth, an initial comparison to other tools or approaches is presented. This will include a comparison to other reviews and studies on the subject of imputing proteomics data. We hope for good discussion on the selected approach and on the envisioned applications of the imputation methods.
Short Abstract: We present Protgraph, a novel approach to represent proteins and digested peptides with feature information, by utilizing a graph structure. It is generated by utilizing the SwissProt-EMBL-Format provided by UniProt, which gives various information about the protein and its features such as isoforms, variants and/or signal peptides. Protgraph maps the canonical sequence to a path in a directed acyclic graph extending it with feature- and digestion-information by adding specific nodes and edges. All possible paths can then be retrieved by traversing from a dedicated start and end node yielding digested peptides. Via a dynamic programming approach different statistics can be retrieved in an efficient manner. We show that the search space in peptide identification increases exponentially when using isoform- and variant-features. With protgraph we provide a fast tool for generating a graph structure from proteins and offer multiple export functionalities, such as FASTAs for search engines, to be used with other tools. As a next step, we would like to explore more sophisticated algorithms on this graph structure to further investigate the search space and to be able to query the graph directly and efficiently for weights.
Short Abstract: Background
Histone post-translational modifications (PTMs) play an important role in our system by regulating the structure of chromatin and therefore contribute to the regulation of gene and protein expression. Irregularities in histone PTMs can lead to a variety of different diseases including various forms of cancer. Histone modifications are analyzed using high resolution mass spectrometry, which generate large amounts of data that requires sophisticated bioinformatics tools for analysis and visualization. PTMViz is designed for downstream differential abundance analysis and visualization of both protein and/or histone modifications.
Results
PTMViz provides users with data tables and visualization plots of significantly differentiated proteins and histone PTMs between two sample groups. All the data is packaged into interactive data tables and graphs using the Shiny platform to help the user explore the results in a fast and efficient manner to assess if changes in the system are due to protein abundance changes or epigenetic changes.
Conclusion
We identified several proteins differentially regulated in the dopaminergic pathway between mice treated with methamphetamine compared to a saline control. We also identified histone post-translational modifications including histone H3K9me, H3K27me3, H4K16ac, and that were regulated due to drug exposure.
