Biology subjects

Seber, P.

Publications and source records attributed to Seber, P..

4 recordsLinked to original sources

bioRxiv2025

Machine-Learning-Based Prediction of Chinese Hamster Ovary Cell Stability Due to Epigenetic Changes

BackgroundChinese hamster ovary (CHO) cells are the main system for producing recombinant protein biopharmaceuticals, but are inherently unstable, affecting their long-term productivity. This cell instability reduces their productivity over time during perfusion operation, which increases the costs of the resulting biopharmaceutical. No models have been published for the prediction of long-term stability. ResultsIn this work, we create the first models for predicting the long-term stability of CHO cells due to changes in chromatin modification levels and methylation. Multilayer perceptrons are the best-performing models, reaching an F1 score of 59.1% and a Matthews correlation coefficient of 19.4%. The models are successful at identifying stable and highly productive CHO cells. Furthermore, Shapley values and interpretable models are used to investigate model coefficients, contributing biological insight to this problem and helping focus future data collection efforts. The models trained in this work are free and open source and available at github.com/PedroSeber/CHO_stability_prediction, allowing their use in the biopharmaceutical industry, reproduction of this work, and the retraining of models on other datasets. ConclusionsWe show that it is possible to train accurate machine learning models to predict the long-term stability of CHO cells using only epigenetic data. The models have high performance and excel in industrially relevant contexts, and thus can improve the bioproduction of medications, especially recombinant proteins. By providing the first predictive models for this task, this work also serves as a foundation for future data collection and modeling efforts.

bioinformatics↗

bioRxiv2024

Improving N-Glycosylation and Biopharmaceutical Production Predictions Using AutoML-Built Residual Hybrid Models

N-glycosylation has many essential biological roles, and is important for biotherapeutics as it can affect drug efficacy, duration of effect, and toxicity. Its importance has motivated the development of mechanistic models for quantitatively predicting the distribution of N-glycans during therapeutic protein production. Here we present a residual hybrid modeling approach that integrates mechanistic modeling with machine learning to produce significantly more accurate predictions for production of monoclonal antibodies in batch, fed-batch, and perfusion cell culture. For the largest dataset, the residual hybrid models have an average 736-fold reduction in testing prediction error. Furthermore, the residual hybrid models have lower prediction errors than the mechanistic models for all of the predicted variables in the datasets. We provide the automatic machine learning software used in this work, allowing other researchers to reproduce this work and use our software for other tasks and datasets.

bioinformatics↗

bioRxiv2023

Recurrent Neural Network-based Prediction of O-GlcNAcylation Sites in Mammalian Proteins

O-GlcNAcylation has the potential to be an important target for therapeutics, but a motif or an algorithm to reliably predict O-GlcNAcylation sites is not available. In spite of the importance of O-GlcNAcylation, current predictive models are insufficient as they fail to generalize, and many are no longer available. This article constructs MLP and RNN models to predict the presence of O-GlcNAcylation sites based on protein sequences. Multiple different datasets are evaluated separately and assessed in terms of strengths and issues. The models trained in this work achieve considerably better metrics than previously published models, with at least a two-fold increase in F1 score relative to previously published models; the specific gains vary depending on the dataset. Within a given dataset, the results are robust to changes in cross-validation and test data as determined by nested validation. The best model achieves an F1 score of 36% (more than 3.5-fold greater than the previous best model) and a Matthews Correlation Coefficient of 35% (more than 4.5-fold greater than the previous best model), and, for the F1 score, 7.6-fold higher than when not using any model. Shapley values are used to interpret the model s predictions and provide biological insight into O-GlcNAcylation.

bioinformatics↗

bioRxiv2023

Linear and Neural Network Models for Predicting N-glycosylation in Chinese Hamster Ovary Cells Based on B4GALT Levels

Glycosylation is an essential modification to proteins that has positive effects, such as improving the half-life of antibodies, and negative effects, such as promoting cancers. Despite the importance of glycosylation, predictive models have been lacking. This article constructs linear and neural network models for the prediction of the distribution of glycans on N-glycosylation sites. The models are trained on data containing normalized B4GALT levels in Chinese Hamster Ovary cells. The ANN models achieve a median prediction error of 1.39%, which is 10-fold smaller than for previously published models, and a narrow error distribution. We also discuss issues with other models reported in the literature. We provide all of the software used in this work, allowing other researchers to reproduce the work and reuse or improve the code in future endeavors.

bioinformatics↗