Machine learning approach informs biology of cancer drug response.

BACKGROUND: The mechanism of action for most cancer drugs is not clear. Large-scale pharmacogenomic cancer cell line datasets offer a rich resource to obtain this knowledge. Here, we present an analysis strategy for revealing biological pathways that contribute to drug response using publicly available pharmacogenomic cancer cell line datasets. METHODS: We present a custom machine-learning based approach for identifying biological pathways involved in cancer drug response. We test the utility of our approach with a pan-cancer analysis of ML210, an inhibitor of GPX4, and a melanoma-focused analysis of inhibitors of BRAFV600. We apply our approach to reveal determinants of drug resistance to microtubule inhibitors. RESULTS: Our method implicated lipid metabolism and Rac1/cytoskeleton signaling in the context of ML210 and BRAF inhibitor response, respectively. These findings are consistent with current knowledge of how these drugs work. For microtubule inhibitors, our approach implicated Notch and Akt signaling as pathways that associated with response. CONCLUSIONS: Our results demonstrate the utility of combining informed feature selection and machine learning algorithms in understanding cancer drug response.

A simplified transposon mutagenesis method to perform phenotypic forward genetic screens in cultured cells.

BACKGROUND: The introduction of genome-wide shRNA and CRISPR libraries has facilitated cell-based screens to identify loss-of-function mutations associated with a phenotype of interest. Approaches to perform analogous gain-of-function screens are less common, although some reports have utilized arrayed viral expression libraries or the CRISPR activation system. However, a variety of technical and logistical challenges make these approaches difficult for many labs to execute. In addition, genome-wide shRNA or CRISPR libraries typically contain of hundreds of thousands of individual engineered elements, and the associated complexity creates issues with replication and reproducibility for these methods. RESULTS: Here we describe a simple, reproducible approach using the SB transposon system to perform phenotypic cell-based genetic screens. This approach employs only three plasmids to perform unbiased, whole-genome transposon mutagenesis. We also describe a ligation-mediated PCR method that can be used in conjunction with the included software tools to map raw sequence data, identify candidate genes associated with phenotypes of interest, and predict the impact of recurrent transposon insertions on candidate gene function. Finally, we demonstrate the high reproducibility of our approach by having three individuals perform independent replicates of a mutagenesis screen to identify drivers of vemurafenib resistance in cultured melanoma cells. CONCLUSIONS: Collectively, our work establishes a facile, adaptable method that can be performed by labs of any size to perform robust, genome-wide screens to identify genes that influence phenotypes of interest.

Understanding cancer drug resistance with Sleeping Beauty functional genomic screens: Application to MAPK inhibition in cutaneous melanoma.

Combined BRAF and MEK inhibition is an effective treatment for BRAF-mutant cutaneous melanoma. However, most patients progress on this treatment due to drug resistance. Here, we applied the Sleeping Beauty transposon system to understand how melanoma evades MAPK inhibition. We found that the specific drug resistance mechanisms differed across melanomas in our genetic screens of five cutaneous melanoma cell lines. While drivers that reactivated MAPK were highly conserved, many others were cell-line specific. One such driver, VAV1, activated a de-differentiated transcriptional program like that of hyperactive RAC1, RAC1P29S. To target this mechanism, we showed that an inhibitor of SRC, saracatinib, blunts the VAV1-induced transcriptional reprogramming. Overall, we highlighted the importance of accounting for melanoma heterogeneity in treating cutaneous melanoma with MAPK inhibitors. Moreover, we demonstrated the utility of the Sleeping Beauty transposon system in understanding cancer drug resistance.

SRC-RAC1 signaling drives drug resistance to BRAF inhibition in de-differentiated cutaneous melanomas.

Rare gain-of-function mutations in RAC1 drive drug resistance to targeted BRAF inhibition in cutaneous melanoma. Here, we show that wildtype RAC1 is a critical driver of growth and drug resistance, but only in a subset of melanomas with elevated markers of de-differentiation. Similarly, SRC inhibition also selectively sensitized de-differentiated melanomas to BRAF inhibition. One possible mechanism may be the suppression of the de-differentiated state, as SRC and RAC1 maintained markers of de-differentiation in human melanoma cells. The functional differences between melanoma subtypes suggest that the clinical management of cutaneous melanoma can be enhanced by the knowledge of differentiation status. To simplify the task of classification, we developed a binary classification strategy based on a small set of ten genes. Using this gene set, we reliably determined the differentiation status previously defined by hundreds of genes. Overall, our study informs strategies that enhance the precision of BRAFi by discovering unique vulnerabilities of the de-differentiated cutaneous melanoma subtype and creating a practical method to resolve differentiation status.

Src-Dependent DBL Family Members Drive Resistance to Vemurafenib in Human Melanoma.

The use of selective BRAF inhibitors (BRAFi) has produced remarkable outcomes for patients with advanced cutaneous melanoma harboring a BRAFV600E mutation. Unfortunately, the majority of patients eventually develop drug-resistant disease. We employed a genetic screening approach to identify gain-of-function mechanisms of BRAFi resistance in two independent melanoma cell lines. Our screens identified both known and unappreciated drivers of BRAFi resistance, including multiple members of the DBL family. Mechanistic studies identified a DBL/RAC1/PAK signaling axis capable of driving resistance to both current and next-generation BRAFis. However, we show that the SRC inhibitor, saracatinib, can block the DBL-driven resistance. Our work highlights the utility of our straightforward genetic screening method in identifying new drug combinations to combat acquired BRAFi resistance. SIGNIFICANCE: A simple, rapid, and flexible genetic screening approach identifies genes that drive resistance to MAPK inhibitors when overexpressed in human melanoma cells.