It is extremely important to improve the accuracy of Cancer diagnosis for several patients and those who are healthy but are tested as well. The study brings up a good architecture of Deep Neural Network(DNN) which is able to infer various properties of biological samples, through multi-task and transfer learning. Computational challenges is one of the reasons for molecular cancer pathology to be limited to small number of biomarkers rather than the whole transcriptome. A low-dimensional latent vector is obtained encoding the whole transcription profile and then mRNA, miRNA expression profiles, tissue and disease is recovered from this vector. The model is employed on mRNA transcription profiles of 10750 clinical samples from 34 classes (one healthy and 33 different type of cancer) from 27 tissues. This method outperform prior works and classifiable machine learning models in predicting tissue-of-origin, normal or disease state and type of cancer. It also correctly classified cancer subtype with 99.4% accuracy.
Input variables : mRNA Profile
Output Variables : Status of sample- Tissue or Cancer
| Statistical | : | Somers D | Accuracy | Precision and Recall | Confusion Matrix | F1 Score | Roc and Auc | Prevalence | Detection Rate | Balanced Accuracy | Cohen's Kappa | Concordance | Gini Coefficent | KS Statistic | Youden's J Index |
| Infrastructure | : | Log Bytes | Logging/User/IAMPolicy | Logging/User/VPN | CPU Utilization | Memory Usage | Error Count | Prediction Count | Prediction Latencies | Private Endpoint Prediction Latencies | Private Endpoint Response Count |
Visit Model : nature.com
Additional links : nature.com
| Model Category | : | Public |
| Date Published | : | September, 2018 |
| Healthcare Domain | : | Provider |
| Code | : | github.com |
| Health Risk Management |
| Disease Detection |