Application of Hallmarks of Cancer Gene Signatures Reveals Novel Energetics-Associated Subtypes with Distinct Survival Outcomes in AML

Background: Acute Myeloid Leukaemia (AML) is a heterogeneous disease with a dramatic variation in patient treatment response. The treatment landscape of AML has remained relatively unchanged for over 30 years, presenting the need to further categorise the disease and it's underpinning molecular subtypes. The re-use of published and validated predictive and prognostic gene signatures provide an invaluable in silico opportunity to uncover the biological mechanisms underpinning treatment response in AML.

Aims: The project aims to identify stable patient groups representative of novel molecular AML subtypes via the development of an automated analytical pipeline utilising validated gene signatures, semi-supervised techniques and survival analysis.

Methods: Primary analysis was performed using a training dataset (AML-OHSU) containing 451 AML patient samples, with validation using a secondary dataset (TCGA-LAML, 200 AML patient samples). Both datasets were processed by Almac Diagnostic's ClaraT software platform which provides a comprehensive overview of tumour profiles via gene expression signatures. An automated statistical analysis pipeline was developed using Consensus Clustering, a method that provides quantitative evidence, via bootstrapping, for determining the number and membership of potential robust clusters within a dataset. This was employed using all (n = 1680) possible combinations of algorithms, distances and linkages and 167 unique gene expression signatures categorised by the 10 Hallmarks of Cancer. Clusters identified as stable were subject to further downstream analysis to identify clinical associations (log-rank p < 0.05, chi square p-value < 0.05 and BH FDR <0.2). To identify gene signatures driving cluster structure and thus survival differences, Kruskal-Wallis testing (p < 0.05) was applied using cluster labels. This was followed up with Dunn's tests (BH FDR < 0.2) to establish signature scores that were significant between all pairwise comparisons of clusters.

Results

The automated clustering pipeline returned a total of 1,314 stable cluster results across all algorithm, distance and linkage parameters for the 10 Hallmarks of Cancer in the AML-OHSU dataset. These stable clusters were further filtered and ranked via log-rank testing (p-value <0.05) returning 134 stable clusters with significant differences in survival outcome. Clusters based on signatures (n = 22) representative of the Energetics Hallmark most frequently occurred throughout results.

Stable clusters with significant survival differences were tested against 32 clinical categorical variables present in the AML-OHSU dataset. In the highest-ranked cluster structure (k = 3, log-rank p-value: 0.033), patients associated with Cluster 3 (best survival) had both a high fusion and NPM1 mutation frequency whereas Cluster 2 (poorest survival) contained the highest number of patients above the age of 65.

Out of the 22 Energetics Hallmark signatures, in n = 16, a significant difference in scores between clusters were identified. Signatures associated with Cholesterol Homeostasis, Myogenesis and Heme Metabolism were higher in the poorer survival cluster and lower in the best survival cluster.

Energetics clusters and signature trends were validated in the TCGA dataset. A significant difference in survival probability (Log rank p-value: 0.019) was again found between stable clusters of the energetics hallmark (K = 3). Similarly, high signature scores in Cholesterol Homeostasis and Myogenesis signatures were again associated with the poorest survival outcome cluster.

Conclusion: We have demonstrated that the novel analytical pipeline developed can aid the discovery of new molecular subtypes in AML associated with prognosis and validate these in an independent dataset. To our knowledge, the energetics hallmark has not been linked with AML prognosis before, suggesting novel biology linked to treatment response.

Kennedy:Almac Diagnostic Services UK: Current Employment.