Estimated limits of organism-specific training for epitope prediction

Abstract

BackgroundThe identification of linear B-cell epitopes remains an important task in the development of vaccines, therapeutic antibodies and several diagnostic tests. Machine learning predictors are trained to flag potential epitope candidates for experimental validation and currently, most predictors are trained as generalist models using large, heterogeneous data sets. Recently, organism-specific training has been shown to improve prediction performance for data-rich organisms. Unfortunately, for most organisms, large volumes of validated epitope data are not yet available. This article investigates the limits of organism-specific training for epitope prediction. It explores the validity of organism-specific training for data-poor organisms by examining how the size of the training data set affects prediction performance. It also compares the performance of organism-specific training under simulated data-poor conditions to that of models trained using traditional large heterogeneous and hybrid data sets.ResultsThis work shows how models trained on small organism-specific data sets can outperform similar models trained on (potentially much larger) heterogeneous and mixed data sets. The results reported indicate that as few as 20 labelled peptides from a given pathogen can be sufficient to generate models that outperform widely-used predictors from the literature, which are trained on heterogeneous data. Models trained using more than about 100 to 150 organism-specific peptides perform consistently better than most generalist models across a wide variety of performance measures, and in some cases can even approach the performance of organism-specific models trained on considerably larger data sets.ConclusionsOrganism-specific training improves linear B-cell epitope prediction performance even in situations when only small training sets are available, which opens new possibilities for the development of bespoke, high-performance predictive models when studying data-poor organisms such as emerging or neglected pathogens.

Phylogeny-aware linear B-cell epitope predictor detects candidate targets for specific immune responses to Monkeypox virus

“…Beyond the detection of potential Monkepox LBCEs, the phylogeny-aware approach used in this work has been shown to provide better results than the current standard tool for LBCE prediction (Bepipred 2.0), reinforcing earlier results that already suggested this to be the case [19]. More importantly, the successful development of model tailored specifically for an emerging global pathogen with almost no LBCE data available shows that this approach is useful even beyond the lower bounds investigated in an earlier work [31], and that utilising information from phylogenetically-related pathogens can be a successful strategy for the development of predictive models tailored for specific groups of pathogens that share a common ancestor. With the current wide availability of considerable computing power even on low-cost and mobile devices, there is clearly scope for a wider use of pathogen-specific models, trained and tuned for the prediction task of interest.…”

“…Although this underperformance may seem superficially at odds with the notion of organism-specific model training, we suggest that the extremely low number of OPXV examples may be insufficient to fit models with good predictive abilities. This finding reinforces results obtained in our preliminary exploration of the limits of organism-specific training: at the very low end of data availability the resulting models may not achieve sufficient generalisation ability [31]. It is also interesting to note that the bespoke models (with the exception of the one trained on all viral epitopes) present lower specificity than Bepipred 2.0, but non-inferior PPV and considerably higher sensitivity and NPV.…”

The rise of taxon-specific epitope predictors

“…Supporting evidence for this assumption was presented in [ 2 ], where epitopes from a number of phylogenetically distant pathogens were found to exhibit clearly distinct patterns in terms of location on a feature space, including the superposition of LBCEs from one pathogen with known non-immunogenic peptides from others. This would compromise the performance of models trained without the use of taxonomic information, and motivated the development of organism-specific models, which were shown to significantly improve performance over generalist models [ 2 , 9 ].…”

Ensuring the Future of Renewable Energy: A Critical Review of Reliability Engineering Applications in Renewable Energy Systems

“…In summary, geothermal energy, with its consistent supply, offers an appealing renewable option. However, careful resource management and advances in drilling technology are necessary to fully realize its potential (Ashford & Campelo, 2021).…”