idw – Informationsdienst Wissenschaft
Nachrichten, Termine, Experten
Nachrichten, Termine, Experten
Computer Science: Publication in PLOS Biology
Transport proteins are responsible for the ongoing movement of substrates into and out of a biological cell. However, it is difficult to determine which substrates a specific protein can transport. Bioinformaticians at Heinrich Heine University Düsseldorf (HHU) have developed a model – called SPOT – which can predict this with a high degree of accuracy using artificial intelligence (AI). They now present their approach, which can be used with arbitrary transport proteins, in the scientific journal PLOS Biology.
Substrates in biological cells need to be continuously transported inwards and outwards across the cell membrane to ensure the survival of the cells and enable them to perform their function. However, not all substrates that move through the body should be allowed to enter the cells. And some of these transport processes need to be controllable so that they only occur at a particular time or under specific conditions in order to trigger a cell function.
The role of these active and specialised transport channels is assumed by so-called transport proteins, or transporters for short, a wide variety of which are integrated into the cell membranes. A transport protein comprises a large number of individual amino acids, which together form a complex three-dimensional structure.
Each transporter is tailored to a specific molecule – the so-called substrate – or a small group of substrates. But which exactly? Researchers are constantly searching for matching transporter-substrate pairs.
Professor Dr Martin Lercher from the research group for Computational Cell Biology and corresponding author of a study, which has now been published in PLOS Biology: “Determining which substrates match which transporters experimentally is difficult. Even determining the three-dimensional structure of a transporter – from which it may be possible to identify the substrates – is a challenge, as the proteins become unstable as soon as they are isolated from the cell membrane.”
“We have chosen a different – AI-based – approach,” says Dr Alexander Kroll, lead author of the study and postdoc in the research group of Professor Lercher. “Our method – which is called SPOT – used more than 8,500 transporter-substrate pairs, which have already been experimentally validated, as a training dataset for a deep learning model.”
To enable a computer to process the transporter proteins and substrate molecules, the bioinformaticians in Düsseldorf first convert the protein sequences and substrate molecules into numerical vectors, which can be processed by AI models. After completion of the learning process, the vector for a new transporter and those for potentially suitable substrates can be entered into the AI system. The model then predicts how likely it is that certain substrates will match the transporter.
Kroll: “We have validated our trained model using an independent test dataset where we also already knew the transporter-substrate pairs. SPOT predicts with an accuracy above 92% whether an arbitrary molecule is a substrate for a specific transporter.”
SPOT thus suggests highly promising substrate candidates. “This enables us to limit the search scope for experimenters to a significant degree, which in turn speeds up the process of identifying which substrate is a definite match for a transporter in the laboratory,” says Professor Lercher, explaining the link between bioinformatic prediction and experimental verification.
Kroll adds: “And this applies for any arbitrary transport protein, not just for limited classes of similar proteins, as is the case in other approaches to date.”
There are various potential application areas for the model. Lercher: “In biotechnology, metabolic pathways can be modified to enable the manufacture of specific products such as biofuels. Or drugs can be tailored to transporters to facilitate their entry into precisely those cells in which they are meant to have an effect.”
Detailed caption:
Schematic diagram of the prediction process: Protein databases provide a dataset with 8,500 experimentally validated transporter-substrate pairs to train the model (top). Transport proteins comprise a sequence of amino acids, which have been converted into vectors by a deep learning model (centre left, in different shades of green). Information about potential substrates is also converted into numerical vectors (centre right, in different shades of yellow). These vectors train a so-called gradient boosting model (ensemble of multiple decision trees) to predict whether the molecule is a substrate for a specific transport protein (bottom). (Fig.: HHU/Alexander Kroll)
Alexander Kroll, Nico Niebuhr, Gregory Butler, Martin J. Lercher. SPOT: a machine learning model that predicts specific substrates for transport proteins. PLOS Biology, 2024.
DOI: 10.1371/journal.pbio.3002807
Schematic diagram of the prediction process. Detailed caption at the end of the press release.
HHU/Alexander Kroll
Merkmale dieser Pressemitteilung:
Journalisten, Wissenschaftler
Biologie, Informationstechnik
überregional
Forschungsergebnisse, Wissenschaftliche Publikationen
Englisch
Schematic diagram of the prediction process. Detailed caption at the end of the press release.
HHU/Alexander Kroll
Sie können Suchbegriffe mit und, oder und / oder nicht verknüpfen, z. B. Philo nicht logie.
Verknüpfungen können Sie mit Klammern voneinander trennen, z. B. (Philo nicht logie) oder (Psycho und logie).
Zusammenhängende Worte werden als Wortgruppe gesucht, wenn Sie sie in Anführungsstriche setzen, z. B. „Bundesrepublik Deutschland“.
Die Erweiterte Suche können Sie auch nutzen, ohne Suchbegriffe einzugeben. Sie orientiert sich dann an den Kriterien, die Sie ausgewählt haben (z. B. nach dem Land oder dem Sachgebiet).
Haben Sie in einer Kategorie kein Kriterium ausgewählt, wird die gesamte Kategorie durchsucht (z.B. alle Sachgebiete oder alle Länder).
