Summary of Path-gptomic: a Balanced Multi-modal Learning Framework For Survival Outcome Prediction, by Hongxiao Wang et al.
Path-GPTOmic: A Balanced Multi-modal Learning Framework for Survival Outcome Prediction
by Hongxiao Wang, Yang Yang, Zhuo Zhao, Pengfei Gu, Nishchal Sapkota, Danny Z. Chen
First submitted to arxiv on: 18 Mar 2024
Categories
- Main: Computer Vision and Pattern Recognition (cs.CV)
- Secondary: Machine Learning (cs.LG); Genomics (q-bio.GN)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary This paper presents a novel framework for predicting cancer survival outcomes by combining pathology images and genomic data. The existing approaches often overlook valuable biological insights and have one modality dominating the optimization process, leading to inadequate training for the other modality. To address these challenges, the authors introduce the “Path-GPTOmic” framework, which first regulates the embedding space of a foundation model scGPT, trained on single-cell RNA-seq data, to make it adaptable for bulk RNA-seq data. The framework also proposes a gradient modulation mechanism tailored to the Cox partial likelihood loss for survival prediction, ensuring that both modalities are sufficiently trained during the training process. The authors evaluate their model on two TCGA datasets and achieve substantially improved survival prediction accuracy. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper helps us better predict how well someone with cancer will survive. Right now, doctors use two kinds of data: pictures of cells (pathology images) and information about the genes inside those cells (genomic data). However, these methods often miss important clues about what makes some cancers more or less deadly. The authors develop a new way to combine these two types of data called “Path-GPTOmic”. It’s like a special tool that helps doctors understand how different genes and cell features are connected. This tool is better at predicting survival rates than current methods. |
Keywords
* Artificial intelligence * Embedding space * Likelihood * Optimization
