ALL-pharmacotyping data portal: An open-source platform for exploring genomics of leukemia drug susceptibility and resistance

Abstract Acute lymphoblastic leukemia (ALL) is the most common childhood cancer, characterized by a multitude of molecular subtypes with vastly different prognoses. However, the biological basis of inter-patient variability in treatment response remains poorly understood. In particular, the exact impact of ALL genomics on leukemia sensitivity to individual chemotherapeutic agents are largely unclear. Consequently, frontline ALL treatment protocols worldwide continue to employ a largely uniform set of anti-leukemic drugs with limited variation in their application. In the era of personalized medicine, it is now imperative to comprehensively map “gene-drug” relationships in ALL and use it to guide the development of the next generation of clinical trials for this disease. To this end, two recent studies from our group have reported foundational insights into the pharmacogenomic landscape of ALL, with adults and children profiled for both drug sensitivity and leukemia genomics (Lee et al., Nat Med 2023; Yoshimura, et al., J Clin Oncol 2024). Nonetheless, much of these rich datasets remain underutilized in their potential for discovery research. Therefore, we sought to develop an open-source platform for exploring ALL pharmacogenomics, with a series of innovative data visualization and analysis tools. This web-based portal, the ALL-Pharmacotyping Data Portal (APDP), is a freely accessible resource encompassing drug sensitivity and genomic data from over 1,000 pediatric and adult ALL cases, i.e., up to 21 drugs per case with a total of 7,975 datapoints. The portal facilitates identification of drug sensitivity patterns across molecular subtypes of ALL, clinical and demographic features, and their correlations with transcriptomes and treatment response. Examples of these analyses are provided below:Compare ALL subtype vs. drug sensitivity: Users can examine how LC50 values vary across 23 different molecular subtypes of ALL using interactive boxplots, violin plots, and statistical tests. This enables quick identification of subtypes with distinct sensitivity patterns to specific drugs.On-the-fly differential expression (DE) analysis of sensitive vs. resistant ALL samples: For any drug, the portal allows users to classify ALL samples as sensitive or resistant based on LC50 thresholds and run DE analysis between the two groups. Ranked DE genes can be exported or used for hierarchical clustering, heatmap visualization, and gene set enrichment analysis which can highlight enriched biological pathways and molecular processes potentially underlying drug sensitivity or resistance.Visualize drug sensitivity on transcriptome UMAP: This enables intuitive visual correlation between drug response and transcriptomic clustering patterns. From any region of the UMAP, users can select clusters of interest and launch DE analysis to further explore differences between sensitive and resistant samples.Multivariable regression analysis: This allows users to confirm the statistical correlation of drug sensitivities with clinical or gene expression variables by adjusting for co-variates including patient demographics such as sex, genetic ancestry, and age at diagnosis. This integrated framework allows investigators to explore pharmacogenomic patterns in ALL from multiple angles, combining statistical rigor, interactive visualization, and intuitive navigation to accelerate discovery of novel insights into drug response biology.