Research activities
The fellow completed specific training for RNA/DNA/protein extractions, processing and analysis of clinical samples in the first 6 months of the project. A previously generated small transcriptomic dataset (13 patients) was analysed while new data was being generated to improve the fellow’s bioinformatics expertise. Relevant tissue blocks had been identified and the process of sectioning and extracting mRNA, DNA and proteins for further analysis was started.
Whole transcriptome data on 83 sequential samples from 30 patients was generated in the second 6 months of the project. A trial reverse-phase protein array (RPPA) experiment has been performed to determine the best method to purify protein. In addition to RPPA, optimization experiments with a more high-throughput approach, protein mass spectroscopy, was also performed in parallel.
Transcriptomics analysis performed using R and BioConductor packages in 12-18 months of the project. Transcriptomics data was pre-processed and successfully combined with previously generated transcriptomic dataset, bringing the sample numbers up to 167. Differentially expressed genes in dormant and acquired resistant patients under long-term treatment have been determined. Proteomics analysis of a total of 62 samples was performed using label-free quantitation method in mass spectrometry.
In the last 6 months of the project, comparative analysis of gene expression and proteomics data from neoadjuvant (letrozole)-treated dormant tumours with recurrence samples under long-term adjuvant therapy has been performed using different approaches. Features significantly differentially expressed in both omics approaches significantly separated a subset of acquired resistant patients and dormant patients both after long-term and early-on (first 4 months) treatment.
Overview and dissemination of the results
This is the first patient-matched gene expression study to look at long-term aromatase inhibitor-induced dormancy and acquired resistance in breast cancer. We have identified transcriptome- and proteome-wide expression profiles in dormant and acquired resistant tumours after extended-neoadjuvant letrozole therapy (Objective 1). The challenge in analysing this unique dataset to identify “predictive molecular signature” of dormancy (Objective 2) was mainly due to the heterogeneity of response and resistance amongst patients. To overcome this, we have utilized machine learning and successfully identified genes that differentiate a subset of acquired resistant tumours from dormant tumours after long-term treatment and also in the first 4 months of therapy. Three peer-reviewed research articles have been published in high-quality journals, one has recently been submitted and another is currently in preparation.
