Analytische Chemie
Filtern
Dokumenttyp
- Posterpräsentation (8) (entfernen)
Sprache
- Englisch (8)
Referierte Publikation
- nein (8)
Schlagworte
- Bioinformatics (4)
- Taxonomic analysis (4)
- Graphical models (3)
- Virus protoemics (3)
- Mass spectrometry (2)
- Algorithms (1)
- Artificial intelligence (1)
- Data privacy (1)
- Deep Learning (1)
- Embedded system (1)
Organisationseinheit der BAM
- VP Vizepräsident (8) (entfernen)
The deployment of machine learning (ML) and deep learning (DL) in structural health monitoring (SHM) faces multiple challenges. Foremost among these is the insufficient availability of extensive high-quality data sets essential for robust training. Within SHM, high-quality data is defined by its accuracy, relevance, and fidelity in representing real-world structural scenarios (pristine as well as damaged). Although methods like data augmentation and creating synthetic data can add to datasets, they frequently sacrifice the authenticity and true representation of the data. Sharing real-world data encapsulating true structural and anomalous scenarios offers promise. However, entities are often reluctant to share raw data, given the potential extraction of sensitive information, leading to trust issues among collaborating entities.
Our study introduces a novel methodology leveraging Federated Learning (FL) to navigate these challenges. Within the FL framework, models are trained in a decentralized manner across different entities, preserving data privacy. In our research, we simulated several scenarios and compared them to traditional local training methods. Employing guided wave (GW) datasets, we distributed the data among different parties (clients) using IID (independent, identically distributed or in other words, statistically identical) mini batches of dataset, as well as non-IID configurations. This approach mirrors real-world data distribution among varied entities, such as hydrogen refueling stations.
In our methodology, the initial round involves individualized training for each client using their unique datasets . Subsequently, the model parameters are sent to the FL server, where they are averaged to construct a global model. In the second round, this global model is disseminated back to the clients to aid in predictive tasks. This iterative process continues for several rounds to reach convergence.
Our findings distinctly highlight the advantages of FL over localized training, evidenced by a marked improvement in prediction accuracy . This research underscores the potential of FL in GW-based SHM, offering a remedy to similar challenges tied to data scarcity in other SHM approaches and paving the way for a new era of collaborative, data-centric monitoring systems.
In mass spectrometry based proteomics, protein homology leads to
many shared peptides within and between species. This complicates
taxonomic inference. inference. We introduce PepGM, a graphical model for taxonomic profiling of viral proteomes and metaproteomic datasets.
Using the graphical model, our approach computes statistically sound
scores for taxa based on peptide scores from a previous database
search, eliminating the need for commonly used heuristics.
In mass spectrometry based proteomics, protein homology leads to
many shared peptides within and between species. This complicates
taxonomic inference. inference. We introduce PepGM, a graphical model for taxonomic profiling of viral proteomes and metaproteomic datasets.
Using the graphical model, our approach computes statistically sound
scores for taxa based on peptide scores from a previous database
search, eliminating the need for commonly used heuristics.
In mass spectrometry based proteomics, protein homology leads to
many shared peptides within and between species. This complicates
taxonomic inference. inference. We introduce PepGM, a graphical model for taxonomic profiling of viral proteomes and metaproteomic datasets.
Using the graphical model, our approach computes statistically sound
scores for taxa based on peptide scores from a previous database
search, eliminating the need for commonly used heuristics. heuristics.
Probability based taxonomic profiling of viral and microbiome samples using PepGM and Unipept
(2022)
In mass spectrometry based proteomics, protein homology leads to many shared peptides within and between species. This complicates
taxonomic inference in samples of unknown taxonomic origin. PepGM uses a graphical model for taxonomic profiling of viral proteomes and
metaproteomic datasets providing taxonomic confidence scores. To build the graphical model, a list of potentially present taxa needs to be
inferred. To this end, we integrate Unipept, which enables the fast querying of potentially present taxa. Together, they allow for taxonomic
inference with statistically sound confidence scores.
We interpret solving the multi-vehicle routing problem as a team Markov game with partially observable costs. For a given set of customers to serve, the playing agents (vehicles) have the common goal to determine the team-optimal agent routes with minimal total cost. Each agent thereby observes only its own cost. Our multi-agent reinforcement learning approach, the so-called multi-agent Neural Rewriter, builds on the single-agent Neural Rewriter to solve the problem by iteratively rewriting solutions. Parallel agent action execution and partial observability require new rewriting rules for the game. We propose the introduction of a so-called pool in the system which serves as a collection point for unvisited nodes. It enables agents to act simultaneously and exchange nodes in a conflict-free manner. We realize limited disclosure of agent-specific costs by only sharing them during learning. During inference, each agents acts decentrally, solely based on its own cost. First empirical results on small problem sizes demonstrate that we reach a performance close to the employed OR-Tools benchmark which operates in the perfect cost information setting.
Introduction:
With the introduction of accurate deep learning predictors, spectral matching applications might experience a renaissance in tandem mass spectrometry (MS/MS) driven proteomics. Deep learning models, e.g., Prosit, predict complete MS/MS spectra from peptide sequences and give the unprecedented ability to accurately predict mass spectra that may arise from any given proteome. However, the amount of spectral data is enormous when querying large search spaces, e.g., metaproteomes composed of many different species.
Current spectral library search software, such as SpectraST, is not equipped to meet run time and memory constraints imposed by such large MS/MS databases, covering several millions of peptide spectrum predictions.
Methods:
Inspired by the fragment index data structure that had been introduced with MSFragger, we implement an efficient peak matching algorithm for computing spectral similarity between query and library spectra. Mistle (Metaproteomic index and spectral library search engine) uses index partitioning and SIMD (Single instruction, multiple data) intrinsics, which greatly improves speed and memory efficiency for searching large spectral libraries. Mistle is written in C++20 and highly parallelized.
Results:
We demonstrate the efficiency of Mistle on two predicted spectral libraries for the lab-assembled microbial communities 9MM and SIHUMIx. Compared to the spectral library search engine SpectraST, Mistle shows a >10-fold runtime improvement and is also faster than msSLASH, which uses locality-sensitive hashing. Although Mistle is slower than MSFragger, Mistle‘s memory footprint is an order of magnitude smaller. Furthermore, we find evidence that the spectral matching approach to predicted libraries identifies peptides with higher precision. Mistle detects peptides not found by database search via MSFragger and in turn uncovers unnoticed false discoveries among their matches.
Conclusion:
In this study, we show that predicted spectral libraries can enhance peptide identification for metaproteomics. Mistle provides the means to efficiently search large-scale spectral libraries, highlighted for the microbiota 9MM and SIHUMIx.
This paper describes a wireless mobile prototype able to perform optical measurements by means of a miniaturized spectrometer for low light analysis, e.g. fluorescent sensors.
Evaluations, calculations, calibration management and result display are performed by a computer or a standard tablet. The device was designed primarily to detect traces of oil in drinking or ground water and for the analyses of crude oils. However, it can also address a wide range of fluorescent sensors. The fast and user-friendly inspection of water quality or oil properties, as well as the adaptability and mobility, make the device attractive for a variety of users. Further application areas could be easily implemented by adapting the optics and the software (database, data processing and calibration plots, etc.).