Predictive Health Modeling

Builds a predictive model
of cancer survival


Cytology

Blast Percent
Bone Marrow Blast Percent
Bone Marrow Promyelocyte Percent
Bone Marrow Myelocyte Percent
Bone Marrow Metamyelocyte Percent
Bone Marrow Band Percent
Bone Marrow Neutrophil Percent
Bone Marrow Eosinophil Percent
Bone Marrow Basophil Percent
Bone Marrow Lymphocyte Percent
Bone Marrow Monocyte Percent
Bone Marrow Prolymphocyte Percent
Bone Marrow Promonocyte Percent
Abnormal Lymphocyte Percent

Mutations

PML-RAR
FLT3
IDH1 R132
IDH1 R140
IDH1 R172
Activating RAS
NPMc

Other

Year Diagnosed
Age Diagnosed
Risk Category



This web application builds a predictive model of cancer survival based on user-selected features.

Select features desired for the model using the checkboxes. Click Submit to generate the model and display predictions. Data for this web application came from The Cancer Genome Atlas for Acute Myeloid Leukemia. Data was extracted from XML files and stored in a Postgres database on this server. This python web application is powered by Flask. The numpy and sklearn packages are used to handle the data and perform linear regression. The psycopg2 package is used to access the database. Graph generated with plotly.

The process of developing this application is described on Matthew Theisen's Data Blog.