Integrative deep learning analysis improves colon adenocarcinoma patient stratification at risk for mortality- dataset 2
Authors/Creators
- 1. The Jackson Laboratory for Genomic Medicine, Farmington, CT, USA
- 2. Department of Pathology, Wayne State University, Detroit, MI, USA
- 3. Department of Pathology, Yale University, New Haven, CT, USA
Description
Background
Colorectal cancers are the fourth most diagnosed cancer and the second leading cancer in number of deaths. Many clinical variables, pathological features, and genomic signatures are associated with patient risk, but reliable patient stratification in the clinic remains a challenging task. Here we assess how image, clinical, and genomic features can be combined to predict risk.
Methods
We developed and evaluated integrative deep learning models combining formalin-fixed, paraffin-embedded (FFPE) whole slide images (WSIs), clinical variables, and mutation signatures to stratify colon adenocarcinoma (COAD) patients based on their risk of mortality. Our models were trained using a dataset of 108 patients from The Cancer Genome Atlas (TCGA), and were externally validated on newly generated dataset from Wayne State University (WSU) of 123 COAD patients and rectal adenocarcinoma (READ) patients in TCGA (N=52).
Findings
We first observe that deep learning models trained on FFPE WSIs of TCGA-COAD separate high-risk (OS<3years, N=38) and low-risk (OS>5years, N=25) patients (AUC=0.81±0.08, 5year survival p<0.0001, 5year relative risk=1.83±0.04) though such models are less effective at predicting overall survival (OS) for moderate-risk (3years<OS<5years, N=45) patients (5year survival p-value=0.5, 5year relative risk=1.05±0.09). We find that our integrative models combining WSIs, clinical variables, and mutation signatures can improve patient stratification for moderate-risk patients (5year survival p<0.0001, 5year relative risk=1.87±0.07). Our integrative model combining image and clinical variables is also effective on an independent pathology dataset (WSU-COAD, N=123) generated by our team (5year survival p<0.0001, 5year relative risk=1.52±0.08), and the TCGA-READ data (5 year survival p<0.0001, 5 year relative risk=1.18±0.17). Our multicenter integrative image and clinical model trained on combined TCGA-COAD and WSU-COAD is effective in predicting risk on TCGA-READ (5year survival p<0.0001, 5year relative risk=1.82±0.13) data. Pathologist review of image-based heatmaps suggests that nuclear size pleomorphism, intense cellularity, and abnormal structures are associated with high-risk, while low-risk regions have more regular and small cells. Quantitative analysis shows high cellularity, high ratios of tumor cells, large tumor nuclei, and low immune infiltration are indicators of high risk tiles.
Interpretation
The improved stratification of colorectal cancer patients from our computational methods can be beneficial for treatment plans and enrollment of patients in clinical trials.
Files
Files
(28.7 GB)
| Name | Size | |
|---|---|---|
|
md5:eadd1ae7b82764d239598446338451ed
|
417.7 MB | Download |
|
md5:07617b44ebcb7269cbbabf016cb74e91
|
1.2 GB | Download |
|
md5:41fc976ebfa059552313cf7167e503a7
|
483.5 MB | Download |
|
md5:48abc33c19e949634cbd554137ed0a7b
|
1.0 GB | Download |
|
md5:1fca60aae0ed9a4e99761751c1016f4a
|
644.2 MB | Download |
|
md5:b7b80a1ed7682be9cd33883419f42153
|
763.6 MB | Download |
|
md5:43871ed78d10bb13b2b466a611301b2e
|
819.6 MB | Download |
|
md5:08abd2bb7fb3d7a45b889b6e06b341d7
|
838.2 MB | Download |
|
md5:14cc1d36c18398aeba8083318d12f1ea
|
597.7 MB | Download |
|
md5:2ef4429d5b846b702bdcc47faec34b51
|
1.0 GB | Download |
|
md5:bda5c7e621c8b9cd3fc6fe3ae59f6739
|
767.0 MB | Download |
|
md5:8eb573fabb8c74ffb25d2fa8b69d13cc
|
1.2 GB | Download |
|
md5:ecbc9da99d96c81ceff0e891947679ba
|
781.9 MB | Download |
|
md5:c0f5b212a57f2732ee4ddab1642fd36d
|
832.3 MB | Download |
|
md5:44a4e1377c0b38dcf56ba2599b50d162
|
985.4 MB | Download |
|
md5:bca1f5c5cdc39ca9f8ce82717319012b
|
1.2 GB | Download |
|
md5:d4c8b598a4d6b59cf49d3935b503db50
|
1.4 GB | Download |
|
md5:1a660acb50633ebc9a3e3243d4948e8f
|
322.0 MB | Download |
|
md5:a52b69e4a0e3790a85e16d08626a64f3
|
1.1 GB | Download |
|
md5:2f44b10f9f097f141076b41aae0e1a8e
|
393.9 MB | Download |
|
md5:44c6db226fe45dfc44ec91e11eb840bb
|
1.1 GB | Download |
|
md5:2c690c24ab4d8d405969f4940870e85d
|
1.3 GB | Download |
|
md5:f20472c91232811570f1ca97ff79cc45
|
1.0 GB | Download |
|
md5:db774d1d0826316da755fdfe66e66cbc
|
811.7 MB | Download |
|
md5:0645b83aee7b7f885cfca70654f492da
|
969.1 MB | Download |
|
md5:e1af437e57cdcd21776e594bbc73d2ec
|
874.4 MB | Download |
|
md5:5d0e25ebd226b28a60ed260f2b73a04f
|
329.7 MB | Download |
|
md5:ce7dc0fe6e79c70f522a78b144fe26b8
|
1.2 GB | Download |
|
md5:b77109402c0efbe7799ca51e214d4c02
|
458.4 MB | Download |
|
md5:d24be72f72da381d19e12b80077be4f5
|
895.9 MB | Download |
|
md5:4a373954105f1d8fec2be9ebf46fa40e
|
852.8 MB | Download |
|
md5:cdb03014ed811757d86115a32d0564c7
|
1.2 GB | Download |
|
md5:ddea9394a40867b1c979bb7b2f65061c
|
826.6 MB | Download |