Published July 24, 2023 | Version v1

Integrative deep learning analysis improves colon adenocarcinoma patient stratification at risk for mortality- dataset 1

  • 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 (74.3 GB)

Name Size
md5:34f8fb3a807bc13e6900c381c2735c7a
753.3 MB Download
md5:7772b4c151677c7fe2d5bfe16cb3c16d
929.2 MB Download
md5:96eaed28fa52b3b2645541ebabd89fe5
697.8 MB Download
md5:2bff80dde7fabf2c9e2d9d2faa10a3d0
698.7 MB Download
md5:211795198df87046db31e6d9333c7597
833.8 MB Download
md5:e43520761bc4f0757715ede55ee70a12
1.0 GB Download
md5:887e1ed7d613ca7deee7bf111bd43bf3
996.9 MB Download
md5:69bf619e0045171476ca3edcebc11224
840.1 MB Download
md5:7c6912e1843bafcda9b46202823bedce
763.1 MB Download
md5:2a65fb6aaa5d0c2f4736a40c0b62991d
633.4 MB Download
md5:0cbd40ea2b6464fa99b9395fc8b30c90
896.1 MB Download
md5:66967b0f1e144b5e6ad99c60d2242269
1.0 GB Download
md5:eade111e98d6d7c1fecfe9d4bd01db6d
1.1 GB Download
md5:8e770ca2289409092848249634c7f814
936.4 MB Download
md5:44e74c6836100e0206b428bb79bb04e3
1.1 GB Download
md5:5184df75a3c52797bd69806e87455296
1.1 GB Download
md5:aea840bf2cbe672f3bd056156443d1e9
1.0 GB Download
md5:6d11c6a9c819390806f10c5625ee26ca
828.4 MB Download
md5:35b8f30f48c6169f5d4a671f7b189d4a
580.4 MB Download
md5:a30f6a92cf9270debb2a8e469e1862fb
692.0 MB Download
md5:84a01c6bc512656ac4b1ba4a0bcddc67
765.3 MB Download
md5:73892fd9261169c909518175b429a063
768.1 MB Download
md5:ab71c5ebee604eb1e7df031d7fa23981
1.2 GB Download
md5:7679b53c75319c84aa4c97a542a9694e
719.9 MB Download
md5:45c9006525785a7f169c9c0fc8a22e1a
571.3 MB Download
md5:382acbf5f67fd7ff2073a7e61454bc47
783.3 MB Download
md5:a7a54027a998532fe9e15067decbc14a
440.9 MB Download
md5:bc3a9a90896b9a1b1817148b94d2cb15
659.1 MB Download
md5:0bc1c458162f974469746b11602c987c
556.0 MB Download
md5:6cd931f9a647877b14807b125219b6fb
714.6 MB Download
md5:495684466e77d3718e572b3b726a29a3
1.0 GB Download
md5:8e3f9be742ed375797e954e2647e1435
984.5 MB Download
md5:1894f54ef7e87489db85f3885da037e3
773.2 MB Download
md5:44981e5c35abf7bf61eefa31da8cdaf2
1.0 GB Download
md5:aa9ca9e98dc3a16a29aaf88952ea23f1
600.6 MB Download
md5:e121d5956ddf9bcde1874c0bbad63b10
646.4 MB Download
md5:8333fd34125247b3ebb51d1a67337e28
786.4 MB Download
md5:0bef0d0a05603ef512421d59e71ec7ea
638.1 MB Download
md5:50b584167625ba78b79ec6e452b57a10
522.0 MB Download
md5:1ab2d213e534aa447cb9291913494292
740.9 MB Download
md5:6b5c4fbb038c778a0739c055b9ca9f2d
805.8 MB Download
md5:81ec87451cd24c3cef9e015924e350bc
298.6 MB Download
md5:35e87ccdb03251d484cf2fbd7ba4e5c1
1.2 GB Download
md5:8d3dbaf2280ac0e0a3759db2e7f293c3
1.2 GB Download
md5:c29badd380c6ebd731585e211fe90d63
720.1 MB Download
md5:cfb6991e730d0d868faa7a68ea28f9f3
910.3 MB Download
md5:27e84535958aa2e3af76caa35c05a56a
736.8 MB Download
md5:aae79c89d19777734d11ad7679bcce37
703.0 MB Download
md5:4102bfe9ba48dcb7b6e0e2800f780d33
915.8 MB Download
md5:7e74278396d84713c1d2baaaa4f8860b
1.1 GB Download
md5:c69140855bc080257ca3485a02b838fe
703.2 MB Download
md5:45185ae836a7fa043e853432e7cce93d
802.3 MB Download
md5:2e4fd66b7c05c56adaecd8eb87ec84d6
1.1 GB Download
md5:2b0596073c42e34b2aa802ca27828d91
815.8 MB Download
md5:23a442667293fb999bcd7345c0067356
885.1 MB Download
md5:1d091e58dbb952cfa6e663a104a0c9cc
640.3 MB Download
md5:8a4f4ceddbc7637004f9fe663ef4f4cd
741.6 MB Download
md5:0e4fa52bca22b731ce535e71e29455ae
1.2 GB Download
md5:0aadf3e5df7356cddb98a1b6f28e2473
794.1 MB Download
md5:b63e748c4089bd840c10f92879c34b8f
506.3 MB Download
md5:32ee44e2974a8c9ce1f59e76ffa90b87
1.3 GB Download
md5:66957cb4db783555fdab991ba18a3735
1.0 GB Download
md5:d41b2a018c5de1c107567f563d30cb54
1.1 GB Download
md5:74f9a1dad733c721872dc13c2d175246
477.0 MB Download
md5:1cf3d24463f4ff4d1e15bb14a029ba69
796.1 MB Download
md5:1a78466a0e4b97c5f217d6ae5cfd98a4
1.0 GB Download
md5:67a90198f24322fa17c4babee158eb12
823.3 MB Download
md5:53f3c6f4de068f4200045419c75d7ea3
835.3 MB Download
md5:b98c292f795cf2de71d2c23c692f1f50
538.3 MB Download
md5:684a7de6d1b7817a0593e89ebafcd0c2
865.6 MB Download
md5:463534216ecaec9a699357be76b509bb
876.1 MB Download
md5:b4a0da38b9b3774ceeee4890ccbc4d41
1.0 GB Download
md5:2023b9e0d8cca3aac2d8d9bf9bc8b8ce
706.8 MB Download
md5:dd34763de47ba1f5bf7cb9d893c06e04
689.8 MB Download
md5:5821cf7ad61a30a277dd217596854759
897.9 MB Download
md5:9df908b4db80bda9b53c45b9e5ff186d
859.7 MB Download
md5:ed98b388456e6d1d97443e0aff40a977
1.1 GB Download
md5:f363e66ec16b479f6944f072bb539bbe
834.0 MB Download
md5:7e9ae3081cc219deef9c405a8e64f636
1.3 GB Download
md5:3a50935adace7688a36fcc74dc42b470
463.9 MB Download
md5:eae906cc0e5d94fa1fdf54f45498d3f5
820.9 MB Download
md5:66b735a0ee17294226113679ea405568
987.4 MB Download
md5:c431feb5291319aada842f0f661acc1c
977.5 MB Download
md5:d9491fc28c6e621a249cf0f2253f6e57
843.4 MB Download
md5:5f2ede7edf5edcdf695a668963da0b7e
666.1 MB Download
md5:9870549f3ab97fe38337b7b01e0e743d
722.8 MB Download
md5:7b1b6608528e27559dbaa12c0bfe15ff
961.5 MB Download
md5:12f4e5525a29cd768beca86aebf35457
856.1 MB Download
md5:34847bf07250a1e230ff66d080095084
882.3 MB Download