mar
Halvtidskontroll - Suze Julia Roostee
Projekt: Image and multi-omics analyses in breast cancer using machine learning
Granskare: Gottfrid Sjödahl, IKVL, Docent och Carl Troein, Forskare, Centre for Environmental and Climate Science (CEC), LU
Abstract
Background
Breast cancer is a heterogeneous disease that affects millions of women worldwide.
Triple-negative breast cancer (TNBC) is a particularly aggressive form of breast cancer that lacks the expression of oestrogen, progesterone, and HER2 receptors, making it resistant to targeted therapies. The identification of reliable biomarkers and the development of accurate models for the diagnosis, prognosis, and treatment of breast cancer is critical for improving patient outcomes. The analysis of histological images and multi-omics data has shown promise for advancing our understanding of TNBC. By utilising machine learning algorithms in the analysis of histological images and multi-omics data, we may be able to develop more accurate models for the diagnosis, prognosis, and treatment of TNBC, ultimately leading to improved patient outcomes.
Methods
In our research, we use machine learning algorithms in the analysis of histological images, multi-omics data (RNA-seq, Whole Genome Sequencing), and clinical variables. For the computational work, we use image processing and machine learning libraries in Python, Snakemake for pipeline handling, and R for any downstream analysis and multi-omics analysis.
Preliminary Results
Project 1
We developed a pipeline for analysing the staining of different markers in tissue micro-arrays (TMAs) in triple-negative breast cancer (TNBC). The extracted cell counts are in line with pathologist scores and gene-expression data. We then show how cell counts could be integrated with the analysis to gain a better understanding of triple-negative breast cancer. We also show that CD3 positive 1 cell counts (T-cells) can stratify invasive disease-free survival in a TNBC cohort, similarly to tumour-infiltrating lymphocytes (TILs).
Project 2
In this project, we are developing a single-sample classifier that can stratify individual samples based on immune expression measured through bulk RNAseq. We have applied an initial classifier developed in an in-house TNBC cohort to an external TNBC validation cohort with promising results.
Significance
By utilising machine learning algorithms in the analysis of histological images and multi-omics data, we may be able to develop more accurate models for the diagnosis, prognosis, and treatment of breast cancer and specifically TNBC, ultimately leading to improved patient outcomes.
Om evenemanget
Plats:
MV B404, rum E24.
Kontakt:
suze_julia [dot] roostee [at] med [dot] lu [dot] se