MelOD: The Melanoma Omics Dashboard for Multimodal Data Exploration

ABSTRACT

We present MelOD (Melanoma Omics Dashboard), a free, web‐based interactive platform integrating preprocessed data from 16 melanoma studies, including eight bulk transcriptomics, six single‐cell RNA‐seq, and two proteomics datasets. MelOD provides user‐friendly visualization and analysis tools, differential expression, dimensionality reduction, clustering, correlation, and survival analysis without requiring local computational resources. Several datasets include annotations for immunotherapy response, facilitating exploration of resistance and response signatures. Built on RShiny with optimized handling of large datasets, MelOD supports real‐time hypothesis generation, cross‐study validation, and community dataset contributions. Freely accessible online, MelOD lowers barriers to multi‐omics research in melanoma and related fields.

Key Points

• MelOD is a free, web‐based interactive dashboard integrating data from 16 melanoma high‐dimensional omics studies.

• The platform supports bulk transcriptomics, single‐cell RNA‐seq, and proteomics data visualization without requiring local computational resources.

• Several datasets include immunotherapy response annotations, enabling exploration of resistance and response signatures.

• MelOD's modular architecture supports community contributions and real‐time cross‐study validation.

MelOD (Melanoma Omics Dashboard) integrates 16 preprocessed melanoma datasets spanning single‐cell RNA‐seq, proteomics, and bulk transcriptomics through tailored pipelines into a unified, freely accessible web platform. Users can interactively explore results through dedicated visualization modules, including UMAP embeddings, cell proportion plots, heatmaps, violin plots, gene correlation matrices, and volcano plots, without requiring local computational resources. MelOD lowers barriers to hypothesis generation and cross‐study validation in melanoma research.

1. Introduction

Melanoma poses a significant global health challenge due to its increasing incidence and high mortality rates (Roky et al. 2024). Early detection and personalized treatment strategies are crucial for improving patient outcomes, and while immune‐checkpoint blockade has transformed treatment in metastatic melanoma, approximately 40% of patients fail to respond or do not achieve durable benefit (Larkin et al. 2019). Advances in high‐throughput omics technologies, including bulk and single‐cell genomics, transcriptomics, and proteomics, have provided valuable insights into the molecular mechanisms underlying melanoma, facilitating the identification of novel biomarkers and therapeutic targets (Smalley et al. 2016). However, the complexity of these datasets and the lack of easily accessible melanoma‐specific tools create significant barriers for researchers without computational expertise to utilize them in their scientific process.

To address these challenges, we developed the Melanoma Omics Dashboard (MelOD), an interactive application designed to explore and visualize melanoma‐specific omics data. MelOD integrates core analytic workflows, including differential expression, dimensionality reduction, and correlation analysis, within an accessible interface. With several datasets annotated for responses to checkpoint blockade, the dashboard enables exploration of signatures predictive of response and resistance. Its modular architecture provides a one‐stop‐shop for real‐time data exploration, offering an enhanced platform for discovery and collaboration among melanoma researchers.

2. Workflow

The MelOD workflow (Figure 1a) provides exploratory analysis and reporting using preprocessed data from 16 melanoma‐focused omics studies. In total, these datasets encompass over 1100 samples, comprising eight bulk transcriptomics, six single‐cell RNA, and two global proteomics studies (Table S1). To ensure consistency and reliability in downstream analysis, each dataset underwent systematic preprocessing tailored to its data type as follows. Bulk transcriptomics data was processed using the Seq‐N‐Slide pipeline (Dolgalev 2024) which utilizes multiple quality control workflows, STAR alignment (Dobin et al. 2013), featureCounts (Liao et al. 2014) and DESeq2 (Love et al. 2014) for gene level quantitation and differential expression analysis. Proteomics data was processed using MaxQuant for normalization and quality control, ensuring accurate protein quantification and identification (Cox and Mann 2008). Lastly, the six single‐cell RNA‐seq datasets were processed using Seurat (Stuart et al. 2019) with filtering of low‐quality cells, normalization, and dimensionality reduction (e.g., UMAP, t‐SNE).

FIGURE 1.

Overview of MelOD features and workflow. (a) Workflow schematic illustrating the preprocessing steps based on data type, including splitting of single‐cell data and uploading to the cloud. Users select datasets, and MelOD optimizes processing and visualization accordingly. (b) Partial screenshot of the Gene Search page, featuring the gene selection box, study selection options, data table, and interactive volcano plot. (c) Partial screenshot of the Transcriptomics/Proteomics analysis page, displaying the gene selection box, survival curves, Principal Component Analysis (PCA) plot, data table, volcano plot, heatmap, and correlation plot visualizations.

The resulting objects were then used for binary comparisons specific to each study, such as brain metastases versus extracranial metastases, combination therapy versus monotherapy responses or protein expression differences in melanoma cells derived from brain versus extracranial metastases. For datasets supporting multiple comparisons, MelOD allows for selection of specific condition comparisons.

MelOD is hosted on shinyapps.io, which allows for simultaneous multi‐user access while ensuring efficient performance. The application features an intuitive sidebar interface that allows researchers to select studies of interest and access tools tailored to each dataset. Processed datasets are hosted on a private and secure cloud infrastructure, allowing users to fetch and load them into MelOD's integrated environment.

MelOD is built in RShiny using several packages, including shinyDashboard for the dashboard framework and plotly for creating interactive visualizations. To optimize RShiny performance with large omics datasets, MelOD implements several technical strategies. Building on Shiny‐cell (Ouyang et al. 2021), our application splits Seurat objects into smaller, more manageable components. Gene expression matrices are stored using a memory‐efficient HDF5 file format, while metadata and dimensionality reduction coordinates are in separate structures. This data architecture enables rapid loading of and interaction with large, high‐dimensional datasets without the need for substantial local computational resources. Note that as performance metrics such as CPU time and data volume are inherently dependent on server configuration and concurrent user load, these values are not directly exposed to end users of the platform.

3. Key Features and Visualizations

For bulk transcriptomics and proteomics data, MelOD provides comprehensive differential expression visualizations, including customizable volcano plots, heatmaps, gene–gene Pearson correlation matrices, PCA plots, and Kaplan–Meier survival curves (when clinical metadata are available) (Figure 1c). A gene search feature allows users to query specific genes across available studies (Figure 1b), providing detailed and interactive results panels and data tables.

For single‐cell RNA‐seq datasets (Figure 2), MelOD provides cell clustering based on low‐dimensional embeddings such as UMAP, gene expression profiles, and differential gene expression data. Users can explore cell clusters, cell type specific marker genes, gene expression patterns, cell proportions, and gene–gene co‐expression with interactive filtering and customization options. Users can also subset cells based on metadata or expression levels, enabling focused analysis across all plots.

FIGURE 2.

Single‐cell analysis page features. Screenshot showcasing the Single‐Cell Analysis page, including subsetting inputs, a data table, cell and gene information panels, gene co‐expression plots, violin plots, proportion plots, and heatmap visualizations.

MelOD supports data export through multiple formats, allowing users to download processed data, publication‐ready figures (png, pdf), and analysis summaries for external use. This functionality enhances the platform's utility for collaborative research by facilitating data sharing and reproducibility. The application employs reactive programming principles in RShiny to ensure responsive user interactions, and its modular framework supports the addition of new datasets as they become available.

In summary, MelOD provides an accessible platform to explore and analyze melanoma omics data through its user‐friendly design and integrated analytical capabilities. By streamlining data exploration and visualization, it supports the simultaneous generation of hypotheses and external validation across multiple datasets, making it a valuable resource not only for the melanoma research community, but also for cancer scientists and developmental biologists interested in gene function and regulatory mechanisms. The platform's comprehensive integration of diverse data types further facilitates cross‐referencing for researchers both inside and outside the melanoma field, potentially accelerating discoveries in cancer biology and related disciplines. MelOD was designed as an expression‐centric resource, focusing on transcriptomic and proteomic modalities amenable to downstream druggability and repurposing analyses. While genomic (DNA‐based) datasets are not currently included, future updates will expand the available datasets and may incorporate genomic data alongside additional analytic features, including spatial transcriptomics and array data analysis. To foster collaboration and maximize MelOD's impact for the broader scientific community, we encourage researchers to submit additional omics datasets for consideration to be included, subject to approval, with a dedicated link on the homepage to facilitate data contribution requests.

Author Contributions

Ipsita Subudhi: conceptualization, validation, supervision. Markus Schober: funding acquisition, supervision, validation. Milad Ibrahim: data curation, formal analysis, validation. Dania Annuar: validation, investigation. Eva Hernando: funding acquisition, supervision, validation. Pietro Berico: validation, formal analysis, data curation. Paul Sastourne‐Haletou: data curation, formal analysis, conceptualization, investigation, methodology, visualization, writing – review and editing, writing – original draft, software, validation. Adam Walker: conceptualization, data curation, visualization, validation. Iman Osman: supervision, funding acquisition, validation. Kelly V. Ruggles: conceptualization, investigation, funding acquisition, writing – review and editing, writing – original draft, validation, supervision, visualization. Alcida Karz: validation. Paola Angulo Salgado: data curation, formal analysis, validation.

Funding

Supported by the NYULH Metastasis Research Network Center (NIH/NCI 1U54CA263001) and the NYU Melanoma SPORE grant (NCI P50 CA225450).

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Table S1: Overview of datasets used in MelOD: Types and comparison details.

Data Availability Statement

MelOD is accessible at https://ruggleslab.shinyapps.io/MelOD/, and the source code is available at https://github.com/ruggleslab/MelOD.