# Explainable Artificial Intelligence for Melanoma Recurrence Prediction via Integrative Modeling of Multiplexed Imaging with Transcriptomics and Histopathology

> **NIH NIH K99** · MASSACHUSETTS GENERAL HOSPITAL · 2024 · $176,040

## Abstract

ABSTRACT
Recurrence from early-stage (stage I/II) primary melanomas, often undetected until symptomatic metastasis,
accounts for most melanoma mortality. Current prognostic tools are inadequate to identify patients at high risk
of recurrence and metastasis at the time of diagnosis, limiting the ability to provide improved surveillance and
personalized treatment. Additionally, biological mechanisms driving early-stage melanoma recurrence remain
understudied, as previous studies have extensively focused on advanced melanomas. I hypothesize that the
states, interactions, and spatial relationships of cells within the melanoma microenvironment, as well as
transcriptomic signatures, play a critical role in early-stage melanoma recurrence and can further inform
predictive models. To test this hypothesis, I propose to leverage recent advances in multiplexed tissue imaging,
spatial transcriptomics, and artificial intelligence to robustly quantify known prognostic factors, discover new
biomarkers, and develop explainable machine-learning models for predicting early-stage melanoma recurrence.
The integration of clinical, histopathologic, and multiplexed imaging single-cell data will facilitate the development
of accurate and reliable prognostic tools, which can be deployed in clinical settings following independent
validation. In Aim 1 (K99), I will develop computational approaches to identify and quantify both known prognostic
features and novel biomarkers through multi-modal analyses of multiplexed imaging and spatial transcriptomic
data. In Aim 2 (K99), using an efficient multi-modal imaging method, I will generate multiplexed and conventional
histopathologic hematoxylin and eosin-stained images of the same tissue section for a large cohort and integrate
them with electronic health records to build interpretable machine-learning models for predicting melanoma
recurrence. In Aim 3 (R00), I will validate the machine learning models in an independent cohort and apply the
pipeline for stage III melanoma recurrence and other cancer settings. My Ph.D. in computer science and ongoing
postdoctoral training in computational biology, genomics, and biomedical informatics put me in a unique position
to accomplish this proposed research. During the K99 phase, I will be supported by an outstanding and
interdisciplinary team of mentors, advisors, and collaborators (Dr. Semenov, Dr. Sorger. Dr. Yu, Dr. Shalek, Dr.
Tsao, Dr. Lian, and Dr. Nemeth) with expertise in all aspects of the proposed research. I will acquire new
knowledge and skills in (a) cancer biology and immunology and (b) computational analyses in multiplexed tissue
imaging and spatial transcriptomics. Together with formal coursework, professional training, and institutional
support from the Massachusetts General Hospital and Harvard Medical School, I will bridge my knowledge gap
in computational cancer biology, professional skills, and leadership vital to transition into an independent position
and establis...

## Key facts

- **NIH application ID:** 10796370
- **Project number:** 1K99CA286966-01
- **Recipient organization:** MASSACHUSETTS GENERAL HOSPITAL
- **Principal Investigator:** Guihong Wan
- **Activity code:** K99 (R01, R21, SBIR, etc.)
- **Funding institute:** NIH
- **Fiscal year:** 2024
- **Award amount:** $176,040
- **Award type:** 1
- **Project period:** 2024-02-09 → 2026-01-31

## Primary source

NIH RePORTER: https://reporter.nih.gov/project-details/10796370

## Citation

> US National Institutes of Health, RePORTER application 10796370, Explainable Artificial Intelligence for Melanoma Recurrence Prediction via Integrative Modeling of Multiplexed Imaging with Transcriptomics and Histopathology (1K99CA286966-01). Retrieved via AI Analytics 2026-05-26 from https://api.ai-analytics.org/grant/nih/10796370. Licensed CC0.

---

*[NIH grants dataset](/datasets/nih-grants) · CC0 1.0*