Biology subjects

Yeung, D.

Publications and source records attributed to Yeung, D..

8 recordsLinked to original sources

bioRxiv2025

Sensitivity of Non-Invasive Motor-Unit-Based Gesture Recognition to Signal Degradation

The information encoded by motor units has been successfully harnessed to establish high-fidelity human-machine interfaces (HMI). However, the sensitivity of these interfaces using high-density surface electromyography decomposition, a prevalent method for observing motor unit behaviour, to signal degradations commonly encountered in practical settings remains unexplored. Here, we investigated the effects of additive white Gaussian noise (WGN), channel loss, and electrode shift on pseudo-real-time MU-driven motion classification. Across six wrist movements by 13 participants, we evaluated the performance of two classifiers: linear discriminant analysis (LDA) and deep neural networks (DNN), under different noise conditions. The results indicate that spatial perturbations, including channel loss and electrode shift, significantly affected classification accuracy, with LDA being more susceptible than DNN. Conversely, under intense signal noise (WGN with 5 dB SNR), LDA outperformed DNN, and its simplicity potentially provides greater robustness in a challenging environment. Hence, application-specific signal processing considerations are required depending on the target HMI application. IMPACT STATEMENTPerturbation severely impairs non-invasive motor-unit-based gesture recognition. High signal fidelity and robust system design are essential for practical human-machine interaction.

bioengineering↗

bioRxiv2025

Effects of Spatial and Signal-Imposed Noises on Motor Unit Decomposition

High-density surface electromyography (HD-sEMG) decomposition offers insights into the neural drive through observation of individual motor units (MUs). However, ensuring that this method remains reliable under real-world signal degradations is crucial for its broader application. Therefore, we investigated the impact of three commonly modeled signal degradations on convolutive blind-source-separation (BSS) MU decomposition. 192 HD-sEMG channels were recorded from the forearm muscles of thirteen healthy participants during six wrist movements. Three broad categories of perturbation were introduced, including additive white Gaussian noise (WGN), channel loss, and electrode shift. These perturbations were chosen to mimic challenges encountered in practice, such as ambient electrical noise, electrode failures, and sensor displacement in order to test the MU decomposition algorithms sensitivity. Then, the effects of perturbations on the quantity and quality of extracted MUs and neural drive estimation were assessed. Under non-perturbed conditions, an average of 179 {+/-} 40 MUs were extracted. Severe global WGN significantly reduced extracted MUs by approximately 81%. In contrast, more localized WGN, or channel loss as high as 15%, and electrode shift had minimal impact, with reductions in the number of MU decomposed being less than 6%. Reconstruction of neural drive through smoothed cumulative spike trains was significantly impaired by global WGN, thus leading to increased root mean square error when compared to conditions, while localized perturbations had negligible effects. Therefore, the BSS-based MU decomposition methods seem to be robust against localized noise, channel loss, and minor electrode shifts but are vulnerable to global additive noise. These results highlight the importance of carefully applying MU decomposition approaches in practical settings, maintaining high SNRs in EMG recordings, and preprocessing noise treatment to specific MU-decomposition needs.

bioengineering↗

bioRxiv2025

Impact of Noise on Deep Learning-Based Pseudo-Online Gesture Recognition with High-Density EMG

Deep neural network (DNN)-based approaches have demonstrated high accuracy in surface electromyography (sEMG)-driven gesture recognition under controlled conditions. However, estimation performance is known to degrade substantially when exposed to perturbations commonly encountered in the real-world. Here, we investigate the impact of typical noises on an autoencoder augmented recurrent neural network gesture estimator driven by engineered features and feature sets. High-density sEMG (HD-sEMG) signals offering rich information particularly suited to DNN-based algorithms were collected from thirteen participants performing wrist movements. Three types of synthetic disturbances were introduced: additive white Gaussian noise (WGN), channel loss, and electrode shift. Results indicate that when using amplitude-based features (specifically, the root mean square value and the mean absolute value), the estimator maintains robust performance under increasing WGN and channel loss, whereas its performance deteriorates markedly with features reflecting signal dynamics and fluctuations, like slope sign changes and zero crossings. Under electrode shift conditions, employing a combined feature set enhances the classifiers resilience. Importantly, the degree of performance degradation depends on both the type and intensity of the noise. These findings confirm the need for noise-resilient architectures in order to achieve practical, everyday sEMG-driven human-machine interfaces. Clinical RelevanceQuantifying the sensitivity of sEMG-based gesture classifiers to noise can help clinicians tailor electrode placement and training and ultimately decrease user frustration and improve the acceptance of assistive devices.

bioengineering↗

bioRxiv2025

Single-cell transcriptomics-guided development of flow cytometric tests predicting chronic myeloid leukemia blast crisis transformation at chronic phase diagnosis

Clinical risk scores in chronic myeloid leukemia (CML) are inadequate for identifying chronic phase (CP) patients at high-risk of blast crisis (BC) progression. The lack of accurate predictive tests hamper timely interventions, including stem cell transplants, that are more effective in early disease. By interrogating a single cell atlas of primary imatinib resistance for a BC-like gene expression signature, we identify aberrant CD42A+ megakaryocytic and CD10+CD19+ lymphoid progenitor expansion, as well as STAT1- and IFN{gamma}-related inflammatory programs, as consistent features in the bone marrow and peripheral blood of CP patients at high-risk of BC transformation. We develop multi-color flow cytometry-based tests (MFC) to detect these features in CP patients at the time of diagnosis. Validating our MFC panels on a combined Australia-Singapore cohort comprising 28 CP patients, including 14 who underwent transformation, we demonstrate their ability to detect 100% of CP patients who transform with no false positives. Our findings highlight small populations of inflamed hematopoietic stem and progenitor cells, present at CP diagnosis, as powerful harbingers of future BC transformation. The ability of MFC panels to detect these cells at diagnosis support their inclusion as accurate risk-assessment tools to improve management of high-risk patients.

cancer biology↗

bioRxiv2024

Zero-shot retention time prediction for unseen post-translational modifications with molecular structure encodings

Mass spectrometry-based proteomics relies on accurate peptide property prediction models to enhance peptide identification and characterization, especially when dealing with peptidoforms. However, current approaches are limited in their ability to generalize to peptides with novel post-translational modifications (PTMs) due to insufficient training data. To address this challenge, we introduce MoSTERT (Molecular Structure Transformer Encoder for Retention Time prediction) and its enhanced variant, MoSTERT-2S, two transformer-based models designed for zero-shot prediction of retention times of peptides with unseen PTMs. Unlike conventional models, MoSTERT encodes peptide residues at the molecular structure level, allowing it to handle diverse PTMs. MoSTERT-2S further improves accuracy by employing a two-step strategy: first predicting the retention time of the unmodified peptide, then estimating the retention time shift induced by the PTMs. Evaluation on an external dataset demonstrates that MoSTERT-2S achieves state-of-the-art performance, reducing prediction errors compared to existing methods. Its ability to accurately predict retention times for peptides with a wide variety of PTMs not seen during training highlights its potential for advancing proteomic workflows analyzing proteoforms and protein modifications.

bioinformatics↗

bioRxiv2023

Adaptive HD-sEMG decomposition: Towards robust real-time decoding of neural drive

Neural interfacing via decomposition of high-density surface electromyography (HD-sEMG) should be robust to signal non-stationarities incurred by changes in joint pose and contraction intensity. We present an adaptive real-time motor unit (MU) decoding algorithm and test it on HD-sEMG collected from the extensor carpi radialis brevis during isometric contractions over a range of wrist angles and contraction intensities. The performance of the algorithm was verified using high-confidence benchmark decompositions derived from concurrently recorded intramuscular electromyography (iEMG). In trials where contraction conditions between the initialization and testing data differed, the adaptive decoding algorithm maintained significantly higher decoding accuracies when compared to static decoding methods. Using gold standard verification techniques, we demonstrate the limitations of filter re-use decoding methods and show the necessity of parameter adaptation to achieve robust neural decoding.

neuroscience↗

bioRxiv2023

Deep learning from harmonized peptide libraries enables retention time prediction of diverse post translational modifications

In proteomics experiments, peptide retention time (RT) is an orthogonal property to fragmentation when assessing detection confidence. Advances in deep learning enable accurate RT prediction for any peptide from sequence alone, including those yet to be experimentally observed. Here we present Chronologer, an open-source software tool for rapid and accurate peptide RT prediction. Using new approaches to harmonize and false-discovery correct across independently collected datasets, Chronologer is built on a massive database with >2.2 million peptides including 10 common post-translational modification (PTM) types. By linking knowledge learned across diverse peptide chemistries, Chronologer predicts RTs with less than two-thirds the error of other deep learning tools. We show how RT for rare PTMs, such as OGlcNAc, can be learned with high accuracy using as few as 10-100 example peptides in newly harmonized datasets. This iteratively updatable workflow enables Chronologer to comprehensively predict RTs for PTM-marked peptides across entire proteomes.

bioinformatics↗

bioRxiv2020

Sequence-specific model for predicting peptide collision cross-section values in proteomic ion mobility spectrometry

The contribution of peptide amino-acid sequence to collision cross-section values (CCS) has been investigated using a dataset of [~]134,000 peptides of four different charge states (1+ to 4+). The migration data was acquired using a two-dimensional LC/trapped ion mobility spectrometry/quadrupole/time-of-flight MS analysis of HeLa cell digests created using 7 different proteases and was converted to CCS values. Following the previously reported modeling approaches using intrinsic size parameters (ISP), we extended this methodology to encode the position of individual residues within a peptide sequence. A generalized prediction model was built by dividing the dataset into 8 groups (four charges for both tryptic/non-tryptic peptides). Position dependent ISPs were independently optimized for the eight subsets of peptides, resulting in prediction accuracy of [~]0.981 for the entire population of peptides. We find that ion mobility is strongly affected by the peptides ability to solvate the positively charged sites. Internal positioning of polar residues and proline leads to decreased CCS values as they improve charge solvation; conversely, this ability decreases with increasing peptide charge due to electrostatic repulsion. Furthermore, higher helical propensity and peptide hydrophobicity result in preferential formation of extended structures with higher than predicted CCS values. Finally, acidic/basic residues exhibit position dependent ISP behaviour consistent with electrostatic interaction with the peptide macro-dipole, which affects the peptide helicity.

bioinformatics↗