Back

Optimized data representation and convolutional neural network model for predicting tumor purity

Sun, G. J.; Jenkins, D. F.; Cingolani, P. E.; Dry, J. R.; Lai, Z.

2019-10-17 bioinformatics
10.1101/805135 bioRxiv
Show abstract

Here we present a machine learning model, Deep Purity (DePuty) that leverages convolutional neural networks to accurately predict tumor purity from next-generation sequencing data from clinical samples without matched normals. As input, our model utilizes SNP-based copy number and minor allele frequency data formulated as a scatterplot image. With a representation matching that used by expert human annotators, we best an existing algorithm using only ~100 manually curated samples. Our simple, data-efficient approach can serve as a straightforward alternative to traditional, more complex statistical methods, for building performant purity prediction models that enable downstream bioinformatic analysis of tumor variants and absolute copy number alterations relevant to cancer genomics.

Matching journals

The top 4 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.