After decades of stagnation, clinical options for acute myeloid leukemia (AML) have expanded in the past few years. Regulatory approval of several new targeted agents by the US Food and Drug Administration — midostaurin for patients with FLT3 mutations, enasidenib for relapsed or refractory patients with IDH2 mutations, and ivosidenib for patients with an IDH1 mutation — has changed the standard of care in AML. These agents provide a mutation-directed treatment approach applicable to as many as 45% of adult patients with AML.
Concurrently, several computational tools have been developed that may accelerate progress in mutation-speciﬁc therapy for AML and other malignancies. In conjunction with comprehensively annotated AML tissue banks, these technical advances have facilitated the creation of large and complex datasets. These include microarray gene expression, exome sequencing, deep sequencing of subclone heterogeneity, RNA sequencing of gene expression, DNA methylation and chromatin, and germline quantitative trait loci.
Making therapeutic decisions based on a patient’s molecular proﬁle is a complex process and requires ongoing, carefully designed preclinical and clinical studies. A continually increasing body of data is available to examine AML heterogeneity and evaluate newer agents for successful clinical development. However, only a limited number of clinicians and experimental hematologists have both the time and the training to access and analyze these repositories. A review article published in Nature summarized currently available data sets and bioinformatic tools that can inform advances in mutation-specific therapy and emphasized the importance of web-based applications that are open, accessible, and user friendly to those without coding experience.
“The future of personalized medicine will likely rely on individualized assessments from multiple types of data, but it will be difficult to conduct clinical trials proving benefit,” said study author Daniel Thomas, MD, PhD, of the Stanford University School of Medicine in California. “For example, if a data approach suggests that AML with a combination of RUNX1, ASXL1, and KRAS is susceptible to a particular set of targeted agents, it will be impossible to conduct a clinical trial to prove this.
“This will be a regulatory challenge for approval and reimbursement that we need to think about and prepare for,” he added. “Ultimately we need to build a database of molecular profiles and response — or lack of response — to targeted therapies so we can help each other learn from other physicians’ experiences.”
Repositories and Searchable Databases
Molecular data have been increasing in diversity and scale, leading to increased applications of machine learning, statistical methodologies, and algorithm development for mining this data. The most recent public data banks that have become available for AML include:
- The Leukemia Gene Atlas (LGA). The LGA is a web interface comprising results from 25 leukemia studies, most of which were in AML. The LGA primarily consists of microarray gene expression proﬁles, but ChIP sequencing, DNA methylation, single nucleotide variant, and patient survival data are also included. When it was first introduced in 2011, the LGA was the only central repository for AML-specific datasets.
- The Cancer Genome Atlas (TCGA) and the Therapeutically Applicable Research to Generate Effective Treatments (TARGET) database. The TCGA is the first major multiomic dataset to be introduced for AML in adults. Included are 200 AML patients with clinical and biospecimen annotations, most of whom also underwent germline sequencing. The TARGET database provides a similar level of assay diversity and clinical annotations for 993 pediatric cases of AML.
- Leucegene and the Minimum Spanning Trees Inferred Clustering (MiSTIC) gene correlation tool. This database differs from the LGA and TCGA in that its web interface is still being expanded. The goal of the Leucegene project is to improve the prognostic classiﬁcation of AML by performing RNA sequencing on 457 AML samples collected from diverse cytogenetic subgroups. This database is supplemented by MiSTIC, an open and available platform designed to help visualize and explore Leucegene RNA sequencing data and other primary cell data.
- BloodSpot. This newly redesigned interactive web portal compares normal and disease states in hematopoiesis. Containing gene expression data from 23 human and murine studies, this portal aims to provide an accessible platform for hematopoietic hypothesis generation.