Periodic Reporting for period 5 - ICON-BIO (Integrated Connectedness for a New Representation of Biology)
Période du rapport: 2023-07-01 au 2025-04-30
The project addressed currently the foremost scientific issues of mining large-scale, heterogeneous, molecular (multi-omic) data to improve biological understanding and medical treatments. It did it by developing sophisticated artificial intelligence (AI) algorithms, which were applied to advance the understanding of uncurable disease and personalize treatment of cancer, rare diseases, Zika epidemic and Covid-19 pandemic. Hence, it is of importance to society.
The overall objectives of the project were achieved. We innovated in computational methods, data science, posed new biological paradigms, and did the applications of foremost importance to society, to help cure serious diseases. In addition to numerous scientific publications and talks, we distributed the output of the project via panel discussion for the non-scientific audiences, as well as writing a graduate level textbook to train new generations of scientists in this important area of research. Seven PhD students defended their PhDs supported by this project. Additional 7 post-docs were trained on this project.
The computational methods for bridging the gap between complex biomedical data, mathematical models and integrative computational analysis methods were advanced, uniting AI and network-science methods, proposing new algorithmic and biological paradigms to help solve the above-mentioned, foremost, real-world problems. In particular, we modeled the multi-scale structure of molecular organization of the data, developed new computational methods to fuse them, and extracted new precision medicine knowledge by applying them. We utilized the best high performance computing (HPC) infrastructure to do this.
Our innovations in data-science and biological paradigms lead to paradigm shifts in data science and biology. In data-science, we displaced the previously dominant paradigm of analyzing one data type in isolation from others. This pioneering work lead towards the new paradigm that is now widely accepted and a subject of continued research in the field of AI, to jointly mine the multi-modal data by “jointly embedding” them in space. In biology, these included our introduction of the new concepts of an “integrated cell” and an “integrated tissue”, models that encompass all available multi-omic data. The project followed the most current bio-technological advancements and developed methods for analyzing their output, including the time-series, single-cell data, obtained by reprogramming of patient-derived cells in health and disease.
The overall objective was to contribute to society by providing new AI methodologies that can utilize the wealth of available bio-medical data to improve personalized medicine (also called precision medicine). We achieved it by designing and applying our new methods to find new genes involved in complex diseases, such as cancer, serious infection (e.g. Covid-19), and rare uncurable diseases, which can further be exploited as biomarkers of disease, or to discover new and better drugs. Also, our methods provided a better sub-typing of patients into sub-groups that should be treated differently. In addition, we re-purposed known drugs to new therapeutic uses, hence reducing the cost of bringing new medications to the market. We did this by developing a new AI methodology that can get evidence coming jointly from all available multi-omic data, which is also explainable and sustainable.
The overall objectives of the project were achieved. We innovated in computational methods, data science, posed new biological paradigms, and did the applications of foremost importance to society, to help cure serious diseases. In addition to numerous scientific publications and talks, we distributed the output of the project via panel discussion for the non-scientific audiences, as well as writing a graduate level textbook to train new generations of scientists in this important area of research. Seven PhD students defended their PhDs supported by this project. Additional 7 post-docs were trained on this project.
The computational methods for bridging the gap between complex biomedical data, mathematical models and integrative computational analysis methods were advanced, uniting AI and network-science methods, proposing new algorithmic and biological paradigms to help solve the above-mentioned, foremost, real-world problems. In particular, we modeled the multi-scale structure of molecular organization of the data, developed new computational methods to fuse them, and extracted new precision medicine knowledge by applying them. We utilized the best high performance computing (HPC) infrastructure to do this.
Our innovations in data-science and biological paradigms lead to paradigm shifts in data science and biology. In data-science, we displaced the previously dominant paradigm of analyzing one data type in isolation from others. This pioneering work lead towards the new paradigm that is now widely accepted and a subject of continued research in the field of AI, to jointly mine the multi-modal data by “jointly embedding” them in space. In biology, these included our introduction of the new concepts of an “integrated cell” and an “integrated tissue”, models that encompass all available multi-omic data. The project followed the most current bio-technological advancements and developed methods for analyzing their output, including the time-series, single-cell data, obtained by reprogramming of patient-derived cells in health and disease.
The overall objective was to contribute to society by providing new AI methodologies that can utilize the wealth of available bio-medical data to improve personalized medicine (also called precision medicine). We achieved it by designing and applying our new methods to find new genes involved in complex diseases, such as cancer, serious infection (e.g. Covid-19), and rare uncurable diseases, which can further be exploited as biomarkers of disease, or to discover new and better drugs. Also, our methods provided a better sub-typing of patients into sub-groups that should be treated differently. In addition, we re-purposed known drugs to new therapeutic uses, hence reducing the cost of bringing new medications to the market. We did this by developing a new AI methodology that can get evidence coming jointly from all available multi-omic data, which is also explainable and sustainable.
We developed new, explainable and sustainable AI methods for mining and extracting new bio-medical knowledge from multi-modal data: systems-level, multi-scale, heterogeneous, molecular and medical (“multi-omics”) data. We achieved our objectives – to improve biological understanding and to contribute to precision medicine. In particular, we developed new AI and network science methods and applied them to multi-omics data for several uncurable diseases, including cancer, Covid-19 and rare diseases. We identified new disease-related genes that could be new biomarkers and proteins to which drugs bind (“drug targets”), we better stratified patients into risk groups that should be treated differently, and we repurposed known drugs to different therapeutics uses and patient groups. The diseases to which our methods were applied include several types of cancer, two rare diseases, and the serious infections including Zika epidemic and Covid-19 pandemic. We also applied it to other domains, e.g. to find genetic determinants of some facial features. Our work is important for society, as it will lead to improving the health and wellbeing of all, also at a reduced cost.
The work was disseminated via publications of 36 refereed scientific journal papers, 3 refereed scientific conference papers, 3 scientific paper pre-prints, 5 refereed book chapters, an edited refereed graduate textbook, 72 invited / keynote talks, 7 contributed talks, software packages, panel discussions and was also covered by the press. The resulting publications are freely available at https://scholar.google.com/citations?user=mLIsLdAAAAAJ(s’ouvre dans une nouvelle fenêtre) . Also, we did a 2-day graduate-level training workshop at Belgrade Bioinformatics Conference (BelBi) 2024.
Technology transfer was extensively explored and lead to a creation of a start-up.
The work was disseminated via publications of 36 refereed scientific journal papers, 3 refereed scientific conference papers, 3 scientific paper pre-prints, 5 refereed book chapters, an edited refereed graduate textbook, 72 invited / keynote talks, 7 contributed talks, software packages, panel discussions and was also covered by the press. The resulting publications are freely available at https://scholar.google.com/citations?user=mLIsLdAAAAAJ(s’ouvre dans une nouvelle fenêtre) . Also, we did a 2-day graduate-level training workshop at Belgrade Bioinformatics Conference (BelBi) 2024.
Technology transfer was extensively explored and lead to a creation of a start-up.
We progressed beyond the state of the art in the following:
1. We provided several abstractions encompassing all available heterogeneous types of omics data. This led us to developing new, explainable and sustainable AI methods for the integration / fusion of the multi-scale, multi-omics data.
2. In addition, we constructed new data science, combinatorial and algebraic topology algorithms for modelling the multi-scale organization of the cellular omics data. They were based on modelling the data by graphs, hypergraphs and abstract simplicial complexes. We included them within our AI methods to improve the analyses results.
3. We furthered the above towards better multi-omic data models based on data embedding. We designed new analytics algorithms based on these.
4. We implemented the above methods and published software packages open source.
5. We applied the above to help cure currently uncurable diseases: various forms of cancer, rare diseases and Covid-19. We advanced several precision medicine applications: biomarker and drug-target discovery, patient sub-typing, and drug re-purposing.
1. We provided several abstractions encompassing all available heterogeneous types of omics data. This led us to developing new, explainable and sustainable AI methods for the integration / fusion of the multi-scale, multi-omics data.
2. In addition, we constructed new data science, combinatorial and algebraic topology algorithms for modelling the multi-scale organization of the cellular omics data. They were based on modelling the data by graphs, hypergraphs and abstract simplicial complexes. We included them within our AI methods to improve the analyses results.
3. We furthered the above towards better multi-omic data models based on data embedding. We designed new analytics algorithms based on these.
4. We implemented the above methods and published software packages open source.
5. We applied the above to help cure currently uncurable diseases: various forms of cancer, rare diseases and Covid-19. We advanced several precision medicine applications: biomarker and drug-target discovery, patient sub-typing, and drug re-purposing.