Inteligentni sistem sklepanja za biološka odkrtija in njegova uporaba pri raziskavah raka
Oznaka in naziv projekta
N2-0236 Inteligentni sistem sklepanja za biološka odkrtija in njegova uporaba pri raziskavah raka
N2-0236 Intelligent inference system for biological discoveries and its application to cancer research
Logotipi ARRS in drugih sofinancerjev
Projektna skupina
Vodja projekta:
Sodelujoče raziskovalne organizacije:
Sestava projektne skupine:
Vsebinski opis projekta
Slovensko:
Zgradili bomo avtonomni sistem, ki bo inteligentno opazoval biološke in klinične vzorce na ravni posameznih celic in sklepal, kako delujejo celični mehanizmi. V zadnjem času partnerja v tem projektnem predlogu razvijata najsodobnejše metode strojnega učenja da bi ugotovili, kako delujejo zapleteni sistemi in olajšali gradnjo sistemov za podporo odločanju, na slovenski strani, in tehnike globokega učenja iz mikroskopskih slik, da bi najbolj natančno izolirali in zbrali posamezne celice in karakterizirali njihove molekularne lastnosti z različnimi metodami, kot so sekvenciranje in masna spektrometrija posameznih celic, na madžarski strani. Za najučinkovitejšo povezavo informacij med večimi vrstami omičnih podatkov bomo skupaj razvili okvir, ki ga sestavljajo slikanje posameznih celic, označevanje in analiza slik z uporabo globokega učenja, izbira celic in sistem sklepanja. Predlagani delotok bo sposoben analizirati biološke probleme, kot je ccRCC (jasnocelični karcinom ledvičnih celic) na ravni posameznih celic, in bo tako zagotovil popolnoma samodejen, nepristranski in človeško razumljiv opis opazovanih vzorcev.
Angleško:
We will build an autonomous system that intelligently observes biological and clinical samples at a single-cell level and infers how cellular mechanisms work. Recently, the partners have been developing state of the art machine learning methods, to infer how complex systems work and to ease the building of decision support systems, at the Slovenian site. Microscopy-based deep learning techniques to most precisely isolate and collect single-cells, and characterize their molecular properties by various methods, including single-cell sequencing and mass spectrometry, has been established at the Hungarian site. To most efficiently connect information between multi-omics modalities, we will develop together a framework that consists of single-cell imaging, annotation and image analysis using deep learning, cell selection and an inference system. The proposed workflow will be capable of analyzing biological problems and will provide a better description of ccRCC (clear cell Renal Cell Carcinoma) at a single-cell level, thus giving a fully automated, unbiased, and human-interpretable description of observed samples.
Osnovni podatki sofinanciranja so dostopni na spletni strani SICRIS.
Delovni sklopi projekta
- WP1. Multimodal data generation.
- WP2. Image analysis and phenotypic discovery using deep learning.
- WP3. Human-interpretable semi-supervised strategies for single-cell selection.