Network-Based Analysis of Exome Sequencing Mutations Identifies Molecular Subtypes of Myelodysplastic Syndromes

Myelodysplastic syndromes (MDS) are a complex, heterogeneous group of disorders characterized by the accumulation of somatic mutations in combinations that vary between patients. While individual mutations have been identified that can risk stratify patients or identify targets for therapies, these findings have been relevant to only a minority of patients, as MDS is a disease not of individual mutations but of combinations of genes acting in molecular networks corresponding to several functional and biological pathways. Further, although two patients with MDS may not have mutations in common, they may share a network affected by similar genes.

We performed exome sequencing of 201 samples from the bone marrow and peripheral blood of patients with MDS, MDS/MPN, and secondary AML (sAML). Network interactions were retrieved from several publically available databases (InACT, MINT, STRING, etc) and uploaded into cytoscape (an open source software platform for visualizing molecular interaction networks and biological pathways). Functional interactions and pathways were uploaded from Reactome and visualized in cytoscape using Reactome Functional Interaction (FI) network function (Reactome WIKI). Survival analyses were calculated from time of diagnosis to last follow up or death on samples with clinical data.

Overall, 3452 mutations were detected with a median of 25 mutations per sample. Network-based analyses identified 745 genes with 293 interactions. Pathway enrichment analysis of the network identified novel pathways that have not been described previously in MDS including: Robo receptor signaling pathway, EphB-Abl signaling pathway, amb2 integrin signaling pathway and NOD-like receptor pathway.  Standard clustering analysis (networks with high connections between nodes within the cluster but sparse connections with nodes in different clusters) identified 6 molecular subtypes of MDS, Figure 1. Pathway enrichment analysis of each subtype identified distinct pathways for each: subtype 1 was enriched mainly in immune mediated pathways, RAS/RAF/MAP kinase signaling pathway, EGFR signaling pathway, VEGF signaling pathway, and ERBB signaling pathway; subtype 2 enriched in spliceosome and RNA polymerase transcription pathways; subtype 3 enriched in mitosis and cell cycle pathways; subtype 4 enriched in cadherin and Wnt signaling pathways; subtype 5 enriched in DNA and histone methylation pathways; and subtype 6 with TP53 and DNA damage pathways. To determine the biological importance of the identified subtypes on outcome, we investigated whether each subtype affected clinical characteristics and overall survival. Overall, clinical data was available for 126 patients. Median age was 70 years, 66% have MDS, 17% MDS/MPN, and 17% s AML, 53% have low risk, 21% intermediate, and 26% high risk by the Revised International Prognostic Scoring System (IPSS-R). Clinical characteristics correlated with molecular subtypes: subtype 6 patients were older (median age 76) with higher blasts percentage (median 7%), 50% had sAML, and 20% RAEB-2 (higher risk by IPSS-R), whereas subtype 3 patients were younger (median age 65), has lower blasts percentage (median 2%) and 83% of them had lower risk MDS by IPSS-R. Excluding samples with overlapping subtypes, the median overall survival for patients with subtype 1,2,3,4,5,6  was 33.0, 24.6, 46.6, 22.9, 25.7, 6.6 months, respectively, p = 0.002. Given similar survival for subtypes 2,4, and 5, these were combined in one group, Figure 1.  To further identify potential genes in our network for targeted therapies, we searched the publically available targeted therapies databases (TARGET and Therapeutic Target Database). We found 30 potential compounds either in clinical trials or under development that could be explored in MDS. 

In conclusion: network-based analyses defined molecular subtypes of MDS that were predictive of survival. It also identified potential targets for novel therapies that are in clinical trials or under development. These subtypes may be useful in the development of precision medicine strategies that are specifically directed at the pathways that are enriched in each subtype.

Figure 1. Network-based analysis subtypes of MDS and overall survival