Adopting a systems approach, we devise a general workflow to define actionable subtypes in human cancers. Applied to small cell lung cancer (SCLC), the workflow identifies four subtypes based on global gene expression patterns and ontologies. Three correspond to known subtypes (SCLC-A, SCLC-N, and SCLC-Y), while the fourth is a previously undescribed ASCL1+ neuroendocrine variant (NEv2, or SCLC-A2). Tumor deconvolution with subtype gene signatures shows that all of the subtypes are detectable in varying proportions in human and mouse tumors. To understand how multiple stable subtypes can arise within a tumor, we infer a network of transcription factors and develop BooleaBayes, a minimally-constrained Boolean rule-fitting approach. In silico perturbations of the network identify master regulators and destabilizers of its attractors. Specific to NEv2, BooleaBayes predicts ELF3 and NR0B1 as master regulators of the subtype, and TCF3 as a master destabilizer. Since the four subtypes exhibit differential drug sensitivity, with NEv2 consistently least sensitive, these findings may lead to actionable therapeutic strategies that consider SCLC intratumoral heterogeneity. Our systems-level approach should generalize to other cancer types. Small-cell lung cancer (SCLC) is an extremely aggressive disease with poor prognosis. Despite significant advances in treatments of other cancer types, therapeutic strategies for SCLC have remained unchanged for decades. We hypothesize that distinct SCLC subtypes with differential drug sensitivities may be responsible for poor treatment outcomes. To this end, we applied a computational pipeline to identify and characterize SCLC subtypes. We found four subtypes, including one (termed “NEv2”) that had not previously been reported. Across a broad panel of drugs, we show that NEv2 is more resistant than other SCLC subtypes, suggesting that this subtype may be partly responsible for poor treatment outcomes. Importantly, we validate the existence of NEv2 cells in both human and mouse tumors. Reprogramming the identity of NEv2 cells into other subtypes may sensitize these cells to existing treatments. However, deciphering global mechanisms that regulate different subtypes is generally unfeasible. To circumvent this, we developed BooleaBayes, a modeling approach that only infers local regulatory mechanisms near stable cell subtypes. Using BooleaBayes, we found master regulators and master destabilizers for each subtype. These findings predict targets that may destabilize a particular subtype, including NEv2, and lead to successful therapy, by either knocking out master regulators or turning on master destabilizers.

Small cell lung cancer (SCLC) is a devastating disease due to its propensity for early invasion and refractory relapse after initial treatment response. Although these aggressive traits have been associated with phenotypic heterogeneity, our understanding of this association remains incomplete. To fill this knowledge gap, we inferred a set of 33 transcription factors (TF) associated with gene signatures of the known neuroendocrine/epithelial (NE) and non-neuroendocrine/mesenchymal-like (ML) SCLC phenotypes. The topology of this SCLC TF network was derived from prior knowledge and was simulated using Boolean modeling. These simulations predicted that the network settles into attractors, or TF expression patterns, that correlate with NE or ML phenotypes, suggesting that TF network dynamics underlie the emergence of heterogeneous SCLC phenotypes. However, several cell lines and patient tumor specimens failed to correlate with either the NE or ML attractors. By flow cytometry, single cells within these cell lines simultaneously expressed surface markers of both NE and ML differentiation, confirming the existence of a “hybrid” phenotype. Upon exposure to standard-of-care cytotoxic drugs or epigenetic modifiers, NE and ML cell populations converged toward the hybrid state, suggesting possible escape from treatment. Our findings indicate that SCLC phenotypic heterogeneity can be specified dynamically by attractor states of a master regulatory TF network. Thus, SCLC heterogeneity may be best understood as states within an epigenetic landscape. Understanding phenotypic transitions within this landscape may provide insights to clinical applications. Cancer Res; 77(5); 1063–74. ©2016 AACR.

Summary: Combinations of multiple pharmacological agents can achieve a substantial benefit over treatment with single agents alone. Combinations that achieve ‘more than the sum of their parts’ are called synergistic. There have been many proposed frameworks to understand and quantify drug combination synergy with different assumptions and domains of applicability. We introduce here synergy, a Python library that (i) implements a broad array of popular synergy models, (ii) provides tools for evaluating confidence intervals and conducting power analysis and (iii) provides standardized tools to analyze and visualize drug combinations and their synergies and antagonisms. Availability and implementation: synergy is available on all operating systems for Python >=3.5. It is freely available from https://pypi.org/project/synergy, and its source code is available at https://github.com/djwooten/synergy. This software is released under the GNU General Public License, version 3.0 or later. Supplementary information: Supplementary data are available at Bioinformatics online.

Whether patients with glioblastoma that contacts the ventricular-subventricular zone stem cell niche (VSVZ + GBM) have a distinct survival profile from VSVZ − GBM patients independent of other known predictors or molecular profiles is unclear. Using multivariate Cox analysis to adjust survival for widely-accepted predictors, hazard ratios (HRs) for overall (OS) and progression free (PFS) survival between VSVZ + GBM and VSVZ − GBM patients were calculated in 170 single-institution patients and 254 patients included in both The Cancer Genome (TCGA) and Imaging (TCIA) atlases. An adjusted, multivariable analysis revealed that VSVZ contact was independently associated with decreased survival in both datasets. TCGA molecular data analyses revealed that VSVZ contact by GBM was independent of mutational, DNA methylation, gene expression, and protein expression signatures in the bulk tumor. Therefore, while survival of GBM patients is independently stratified by VSVZ contact, with VSVZ + GBM patients displaying a poor prognosis, the VSVZ + GBMs do not possess a distinct molecular signature at the bulk sample level. Focused examination of the interplay between the VSVZ microenvironment and subsets of GBM cells proximal to this region is warranted.