Access and use the European prediction service for biological data
Author(s): Ludwig Lautenbacher,Wassim Gabriel,Tobias Schmidt,Marco Schmidt,Dulguun Bold,Christian Panse,Tobias Kockmann,Mathias Wilheim
Affiliation(s): FGCZ ETHZ|UZH
Social media: https://twitter.com/hb9feb
The DLOmix-serving is an open-source and modular machine learning (ML) inference server for biological data based on NVIDIA Triton [1]. The idea is to implement a standardized interface for accessing various prediction models, see [2, 3]. Furthermore, it can be hosted on different institutional sites using a standard Docker image for service robustness and throughput matters. Having the service in place, we expect increased usability and reproducibility of the prediction models. Here we show an example of using the services with the R ecosystem to predict retention time and fragment ions in proteomics experiments. We demonstrate how the models can be accessed from R using HTTP or the Triton client C++ library via Rcpp. Finally, we want to discuss the best way to integrate the client side into the Bioconductor landscape. [1] https://github.com/eubic/EuBIC2023/issues/12 [2] Gessulat, S., Schmidt, T., Zolg, D. P., Samaras, P., Schnatbaum, K., Zerweck, J., ... & Wilhelm, M. (2019). Prosit: proteome-wide prediction of peptide tandem mass spectra by deep learning. Nature methods, 16(6), 509-518. [3] Bouwmeester, R., Gabriels, R., Hulstaert, N. et al. DeepLC can predict retention times for peptides that carry as-yet unseen modifications. Nat Methods 18, 1363–1369 (2021). https://doi.org/10.1038/s41592-021-01301-5