ASMS – 72nd Conference on Mass Spectrometry and Allied Topics

Richard Lee

Introduction

As mass spectrometry innovations continue—introducing higher resolution instruments with more complex data acquisitions—the result is larger data file sizes that introduce a significant computational burden. In addition, user expectations to connect related datasets have increased data complexity, for example in high throughput experimentation (HTE).  Unfortunately, software systems that can account for the increase in data files sizes and data complexity have not advanced at the same rate, predominately remaining as desktop applications that struggle with ever-increasing file sizes.  As informatics systems are transitioning to the cloud and browser-based applications, new technologies are required to handle these large, highly dense data files.  Here, we discuss new technology for facile data handling within a browser-based application.

Methods

A new high density data storage and high performant data processing server have been developed to address the increasing needs of users to process large sets of related data or high resolution data LC/UV/MS data in the cloud.  The data storage is based on HDF5 technology enabling fast I/O that is required for a high performant solution.  The data processing server was developed with cloud-in-mind, and can be scaled according to user load when deployed in a cluster using containerized technology.

Preliminary Data or Plenary Speakers Abstract

A new data storage system, Arcus, was developed based on HDF5 technology, which allows for fast I/O operations.  Arcus allows for a variety of data assemblies or hierarchies to be supported, either as single or multiple related data file acquisitions that can be curated into a single data assembly.  Arcus was initially developed to support HTE workflows, where this technology was used to reduce data processing times from upwards of 45 min to sub 6 min for a 96 well plate of low resolution LC/UV/MS data. Moreover, data from HTE studies are grouped together into a single data assembly to retain relationships between HTE studies and data.

High resolution data has its own challenges, especially when attempting to access and process data on the web due to large file sizes and complex algorithms often used to manipulate the data.  To address these hurdles, Arcus was leveraged to store high density, high resolution LC/UV/MS datasets, and connected to a new processing server.  The unique data processing server was developed with cloud-in-mind.  The high-performance processing server encapsulates advanced processing algorithms and functions and can be deployed using containerization.  Using clustering technology, the processing server can be scaled on-demand based on load.  The front-end interface is browser-based application allowing the user to visualize all related chromatograms, spectra (including MSn spectra), and all relevant meta data. The web application provides a full data processing experience including peak detection, baseline correction, integration, and the ability to add or draw chemical structures to provide chemical context to the data.

Authors: Anne Marie Smith, Alexander Lishchuk, Alexander Sakharov

Introduction 

Continual improvements in MS instruments’ sensitivity and resolution produce highly complex and detailed data. This data requires expert analysis, which can take a significant amount of time. Identification of components can be challenging for complex samples, whether for stability studies, forced degradation, impurity studies, or other areas. These challenges include component co-elution or lack of resolution in the chromatographic peaks. Current standard practice involves a library search to obtain a confident identification of a component. Here, we show a solution that can be applied to various analysis types to accelerate the identification of related sample components without the use of a library.

Methods

The system automatically selects the data files and a list of predefined chemical structures. Background subtraction is optionally performed to leave substances unique to the sample. If a structure list was provided, structures are associated with found components.

The principle structure is fragmented using in-silico prediction. The top fragments along with common modifications of these fragments (i.e., +/-CH2) are searched across MS2 component spectra. Numerical Markush structures are generated for components that contain common ions of the principle to help elucidate the structure more quickly.

Preliminary Data 

The new algorithm uses common ions to identify related components. This helps to identify components that were not part of the provided chemical structure list and gives additional confidence to those that were part of the structure list.

Identification of top in-silico generated principle fragments and their modifications across the MS2 spectra of other components gives the user confidence that the component was related to the principle structure. These components are assigned a numerical Markush, aiding the user in elucidating the component structure. The use of common ions provides the user with insight into the partial component structure and helps to find additional possible components that could otherwise be missed on manual review.

The results are tabulated to summarize all found components, categorizing them as true unknown components (no structure), Markush components (those with a common ion), or confirmed structures (from the structure list). The quality of the structural assignment is based on the presence of common ions, confirmed structure from the structure list, the presence of a confirmed adduct, etc. Users can quickly review components with confidence and focus on true unknowns. All processed data (including mass spectra, chromatograms, and summary tables) is automatically stored in a database and easily retrievable for additional reprocessing. Collaboration and reporting are made easier with multi-user access to the databases and customizable reports.

Novel aspect 

Utilizing common ions between principle structure and related components for accelerated elucidation of unknowns in complex samples.

Authors: Ryan Andrews, Anne Marie Smith, Richard Lee, Sofya Chudova, Vitaly Lashin, Rostislav Pol

Introduction

The demand for more efficient analysis of complex samples is a driving factor of the evolution of technology for analyzing mass spectrometry data. We previously introduced the first vendor-neutral web-based analytical data processing tool, intended for routine, simple analysis of low-resolution LC/UV/MS data. Here, we present an expansion of browser-based software to analyze complex mixtures using an advanced deformulation algorithm. This solution allows low-resolution and high-resolution xC/UV/MS data to be processed, stored, and databased in the same interface—resulting in greater accessibility and flexibility so that scientists are no longer tied to the instrument or lab.

Methods

The user acquires high-resolution xC/UV/MS analytical data from major instrument vendors, bringing it together to process and analyze in one place. Several datasets were run through the advanced component analysis algorithm. The algorithm uses untargeted extraction of chemical features through peak detection, ion grouping, and component interpretation. It can also be used in a targeted manner when supplied with structure(s) and/or mass(es). Users can configure the interface to suit their workflow and scale the processing server based on the system load. Flexibility in deployment (cloud-based, on-premises) allows users to choose a model that best fits their requirements.

Preliminary Data

The component identification process involves the extraction of ion chromatograms (XICs), peak integration, and grouping of spectral features to generate a component mass spectrum within the web browser. Users can also integrate and view XICs and UV slices from the datasets, or view and interpret the mass spectra.

Results from the component extraction algorithm include extracted components (including those that co-elute), pure component spectra, and component MS2 spectra, along with spectral interpretation (molecular, confirmatory, and fragment ions). The algorithm reduces interpretation time and allows the identification of small components that may be overlooked in manual assessments.

Results are stored in a centralized cloud-based database with chemical structure, user interpretations, and metadata to preserve knowledge. Databases are easily searchable by chromatographic, and mass spectral features and data can be retrieved for seamless review or re-analysis through the web-based application. Users can build reports or create templates for routine processes in the pages workspace and view the reports instantly.