Description

Early cancer diagnosis from bisulfite-treated cell-free DNA (cfDNA) fragments require tedious data analytical procedures. Here, we present a Deep-learning-based approach for Early Cancer Interception and DIAgnosis (DECIDIA) that can achieve accurate cancer diagnosis exclusively from bisulfite-treated cfDNA sequencing fragments. DECIDIA relies on feature representation learning of DNA fragments and weakly supervised learning for classification. DECIDIA achieved high classification performance in diagnosis of cancer and cancer-type classification without the requirement of reads mapping.

We provided DECIDIA for diagnosis of colorectal cancer (CRC) and cancer-type prediction (i.e. colorectal cancer vs. hepatocellular carcinoma vs. lung cancer). DECIDIA expects a text input file of sequencing reads one per line. For example, you can put the following reads in a text file read.txt and upload the file. DECIDIA runs prediction on your local browser by using onnxruntime-web. Therefore, it's safe that data never leaves the device for inferencing. It will take a few seconds to cache the model locally for the first time.

ACGAGAAAAACTAACCCCCACCTCCCTCCTAAATAAAATAACTACCAAACAAAAACACTCCTCACTTCCCAAATAAAA
ACATTGAACTTAAAAACAAATACAATTCCCAAACATCTAAACCAAACCACTTTCACCACTACACAACCAAAAATATAC
TTCCTTACAAAAATACTCCCCACATCTCAAATAATAAACGACCGAACAAAAACGCTCCTCACTTCCCAAACAAAACGT

Classifier

Upload A File

Predictions