You are here

Afeez Sodeinde

Afeez Sodeinde

Afeez Sodeinde, a junior biology major at Bemidji State University, used a protein expression dataset to predict cancer dependencies.

Despite advances in our ability to treat cancer and increasing knowledge of cancer biology, we have difficulty identifying the key molecular events that cancer cells depend on for survival. Thus, we need improved tools to identify cancer dependencies, defined as genes that are important in cancer cell survival.
Being at the Broad this summer has been such a fulfilling experience for me. I was surrounded by people who are passionate about science and devoted to helping each other grow. I learned that there are no boundaries to discovery, and that collaborating with others is crucial to solving big problems. Now, I leave with the confidence, exhilaration, and a stronger sense of my capabilities as a scientist! Previous works have suggested that protein expression has been a useful marker and predictor for cancer phenotypes and could potentially predict dependencies. Thus, we hypothesize that we can use protein expression databases to predict cancer cell dependencies. To test this hypothesis, we used a Reverse Phase Protein Array (RPPA) database to determine if protein expression and protein activation could predict cancer dependencies in our CRISPR-Cas9 and shRNA datasets. To identify significant correlations we performed pairwise tests between the cancer dependency (~15,000 genes) and the RPPA datasets (~750 proteins). We simultaneously performed ATLANTIS, a computational program which defines cancer dependencies as a set of proteomic features. We identified several significant dependencies based on imputed features such as protein expression. ATLANTIS identified CDK1 expression best predicted SERBP1 dependency and ERα expression best predicted KDM1A dependency. We designed novel sgRNAs targeting these genes and have verified that these predicted proteins identified cancer dependencies within these cell lines. We further validated these dependencies using flow cytometry and measured proliferation, apoptosis and total cell death using CFSE, Caspase-3 and Annexin-5 respectively. Together, this data validates these genes are predicted dependencies and is an important step forward in identifying cancer vulnerabilities.


Project: Using protein expression to predict cancer dependencies

Mentor: Colles Price, Cancer Program