How does this help cancer?

Machine learning can be used to perform unsupervised learning on the DNA sequences of patient tumors in order to identify non-linear features of the tumor DNA that help predict treatment. Convolutional neural networks (CNN) have led to substantial advances in both supervised and unsupervised learning tasks in computer vision. It is easy to treat DNA sequence (an ordered sequence of A’s, C’s, G’s, and T’s) as an “image”, and therefore apply many of the same techniques used in imaging to DNA sequence. The R package pmsignature extracts characteristic mutation patterns from sets of mutations, as proposed in the paper: A simple model-based approach to inferring and visualizing cancer mutation signatures by Shiraishi, et. al. (2015).

We use Principal Component Analysis (PCA) on these signatures generated by pmsignature, to compute scatter-plots of labelled training data; The DNN is then trained over a large amount of such signatures, ultimately computing and displaying the probability distribution on the heatmap in the output - the distribution of predictions to be made for testing data, having labelled the regions.

We’ve also provided some controls below to enable you tailor the playground to a specific topic or lesson. Just choose which features you’d like to be visible below then save this link, or refresh the page.

What Do All the Colors Mean?

Orange and blue are used throughout the visualization in slightly different ways, but in general orange shows negative values while blue shows positive values.

The data points (represented by small circles) are initially colored orange or blue, which correspond to positive one and negative one.

The blue dots represent labelled BRCA tumors, whereas the orange dots represent the labelled Lung Adenocarcinoma (lungad_10) tumors.

In the hidden layers, the lines are colored by the weights of the connections between neurons. Blue shows a positive weight, which means the network is using that output of the neuron as given. An orange line shows that the network is assigning a negative weight.

In the output layer, the dots are colored orange or blue depending on their original values. The background color shows what the network is predicting for a particular area. The intensity of the color shows how confident that prediction is.

Credits

Created by

| Kumud Ravisankaran | Valeria Brewer | Ninad Mehta | Suraj Jena |

Repurposed the original TensorFlow Playground made by
Daniel Smilkov and Shan Carter under the Apache-2.0 License.
Big Picture and Google Brain teams for feedback and guidance on Neural Networks and the TensorFlow API.