As part of an interdisciplinary project, researchers from our Faculty will prepare an AI architecture and computing infrastructure that will assist in the analysis of three-dimensional models of the human liver. The new tool is designed to support the diagnosis of metabolic diseases, recognizing features such as hypoxia, glycation, and tissue steatosis.

The project “Intelligent agents in precise medicine:from a liver organoid to a predictive AI model in the diagnosis of metabolic diseases,” received funding of nearly 1.9 million PLN under the second edition of our university's “Support for research teams” program.

The project will include 22 researchers from five Wrocław Tech faculties. The project leader is Prof. Aleksandra Kuzan from the Faculty of Medicine.

AI will recognize the “metabolic fingerprint”

The AI architecture and computing infrastructure will be developed by teams from the Faculty of Information and Communication Technology and the Faculty of Management. The work is carried out under the supervision of Prof. Grzegorz Kołaczek, Prof. Rafał Weron, and Prof. Adam Wasilewski. Our Faculty's team represents the field of technical computer science and telecommunications.

Results of biochemical, metabolomic, and lipidomic analyses, as well as quantitative imaging, will be used to develop a tool supporting diagnostic processes. Artificial intelligence will be used not only to organize the gathered information, but also to discover hidden dependencies and patterns.

The researchers will be looking for the so-called “metabolic fingerprint”, a set of features indicating the state of the examined tissue and the disease processes taking place within it.

“The most challenging task is the so-called multimodal fusion, that is combining very different data sources – such as tissue molecular profiles and digital biomarkers from imaging – into a single predictive model capable of recognizing the “metabolic footprint” of the sample,” explains Prof. Grzegorz Kołaczek from the Department of Computer Science and Systems Engineering.

Working with data where the number of analyzed features goes in thousands while the number of biological samples is relatively small will be a particular challenge.

“The 'small-scale regime' requires careful selection of machine learning methods and rigorous validation to prevent overfitting, where the model memorizes individual cases instead of learning general regularities,” explains Prof. Kołaczek.

The explainability of the developed models will be also important. AI should not only determine the tissue condition, indicating characteristics such as hypoxia, glycation, or steatosis, but also show which molecular and imaging features influenced the outcome. Prof. Kołaczek explains that this is an essential condition for building trust in tools that utilize artificial intelligence in medicine.

Due to the sensitive nature of biological and medical data, the entire computing infrastructure, including the AI computing server and graphics accelerator, will be maintained locally at Wrocław University of Science and Technology. This will allow full the researchers to keep full control over the security of information and avoid sharing it outside the university.

Miniature organs in service of science

Data analysed by AI specialists will come from research on liver organoids. Organoids are three-dimensional structures formed from stem cells, which can self-organize under appropriate conditions. They are not full-size organs, but they can replicate selected features of their structure and functioning.

Such structures can demonstrate more effectively than classical cell cultures what happens in tissue under the influence of disease, medication, or altered metabolic conditions.

“In traditional 2D cultures, we deal with a layer of cells of the same type, and they do not reflect the complexity of the human body. Spheroids, on the other hand, although three-dimensional, are typically monocultures. Liver organoids developed in our project will contain not only hepatocytes, but a whole range of cells forming functional tissue that responds physiologically. As a result, this model simulates the functioning of a real organ much better than any previous laboratory solutions,” says Prof. Aleksandra Kuzan from the Department of Preclinical Sciences, Pharmacology and Medical Diagnostics.

Why do we need 3D models?

Over the years, numerous studies have been conducted on 2D cultures, that is on cells growing on a flat surface. Such models are useful, but they have significant limitations. Tissues in the human body are not flat, and cells communicate spatially with each other, responding among others to the proximity of other cell types, substance flow, and local environmental changes.

Therefore, transitioning from traditional cultures to 3D models is an important step towards more effective diagnostics and therapy testing.

“Standard ‘flat’ models involve a significant risk of generating false positive results. For example, many drugs that appear to kill cancer cells in a 2D model even at low concentrations completely fail when faced with a complex 3D structure, which natural tissue resistance. Therefore, in our project we aim at realistic representations, in order to avoid false conclusions already at the stage of basic research,” explains Prof. Aleksandra Kuzan.

The researchers will focus on the liver, as it is a key organ for metabolism, while being susceptible to many different diseases. They will model the processes underlying its disorders, such as inflammation or fibrosis, occurring in diseases related to improper diet, alcohol abuse, genetic predispositions, and metabolic disorders.

The team will also analyse glycation processes, which accompany many metabolic disorders, as well as hypoxic conditions, that is tissue oxygen deprivation. Such a phenomenon may occur, for example, inside neoplastic tumors.

Organoids reflecting various metabolic disorders will also serve to test phytopharmaceuticals, i.e. plant-derived compounds. Researchers will check if they can produce the desired metabolic effect.

Five faculties and one research platform

The project brings together the expertise of researchers from five faculties at Wrocław Tech.

Teams from the Faculty of Information and Communication Technology and the Faculty of Management will prepare the AI architecture and computing infrastructure needed for data analysis.

The Medical Faculty team is responsible for the cultivation of organoids and spheroids as well as basic biochemical analyses. Researchers from the Faculty of Chemistry, under the supervision of Dr. Jerzy Wiśniewski, will conduct advanced lipidomic and metabolomic analyses.

A team from the Faculty of Fundamental Problems of Technology, coordinated by Prof. Igor Buzalewicz, will be responsible for the quantitative imaging of spheroids, organoids, and tissues. By using various sources of optical contrast, the researchers will be able to identify digital fibrosis biomarkers. These are measurable characteristics of the image showing, among others, changes in tissue structure, its heterogeneity, and the degree of collagen accumulation and organization.

“The most challenging stage of the project will be the actual derivation of the cell line from stem cells so that they perfectly differentiate into the desired types of liver cells. In this process, we will collaborate with Prof. Aleksandra Pawlak from the University of Georgia, a world-class expert in the field of organoids,” adds Prof. Aleksandra Kuzan.

The project will also be an opportunity for students from the Histology Research Club, who will participate for example in preparing the 2D cultures.