Drug testing of monodisperse arrays of live microdissected tumors using a valved multiwell microfluidic platform.

Cancer drug testing in animals is an extremely poor predictor of the drug's safety and efficacy observed in humans. Hence there is a pressing need for functional testing platforms that better predict traditional and immunotherapy responses in human, live tumor tissue or tissue constructs, and at the same time are compatible with the use of mouse tumor tissue to facilitate building more accurate disease models. Since many cancer drug actions rely on mechanisms that depend on the tumor microenvironment (TME), such platforms should also retain as much of the native TME as possible. Additionally, platforms based on miniaturization technologies are desirable to reduce animal use and sensitivity to human tissue scarcity. Present high-throughput testing platforms that have some of these features, e.g. based on patient-derived tumor organoids, require a growth step that alters the TME. On the other hand, microdissected tumors (µDTs) or "spheroids" that retain an intact TME have shown promising responses to immunomodulators acting on native immune cells. However, difficult tissue handling after microdissection has reduced the throughput of drug testing on µDTs, thereby constraining the inherent advantages of producing numerous TME-preserving units of tissue for drug testing. Here we demonstrate a microfluidic 96-well platform designed for drug treatment of hundreds of similarly-sized, cuboidal µDTs ("cuboids") produced from a single tumor sample. The platform organizes a monodisperse array of four cuboids per well in 384 hydrodynamic traps. The microfluidic device, entirely fabricated in thermoplastics, features 96 microvalves that fluidically isolate each well after the cuboid loading step for straightforward multi-drug testing. Since our platform makes the most of scarce tumor tissue, it can potentially be applied to human biopsies that preserve the human TME while minimizing animal testing.

Microdissected tumor cuboids: a microscale cancer model that retains a complex tumor microenvironment.

Present cancer disease models - typically based on cell cultures and animal models that lack the human tumor microenvironment (TME) - are extremely poor predictors of human disease outcomes. Microscale cancer models that combine the micromanipulation of tissues and fluids offer the exciting possibility of miniaturizing the drug testing workflow, enabling inexpensive, more efficient tests of high clinical biomimicry that maximize the use of scarce human tissue and minimize animal testing. Critically, these microscale models allow for precisely addressing the impact of the structural features of the heterogeneous TME to properly target and understand the contributions of these unique zones to therapeutic response. We have recently developed a precision slicing method that yields large numbers of cuboidal micro-tissues ("cuboids", ∼ (400 µm) 3 ) from a single tumor biopsy. Here we evaluate cuboids from syngeneic mouse tumor models and human tumors, which contain native immune cells, as models for drug and immunotherapy evaluation. We characterize relevant TME parameters, such as their cellular architecture (immune cells and vasculature), cytokine secretion, proteomics profiles, and their response to drug panels in multi-well arrays. Despite the cutting procedure and the time spent in culture (up to 7 days), the cuboids display strong functional responses such as cytokine and drug responses. Overall, our results suggest that cuboids make an excellent model for applications that require the TME, such as immunotherapy drug evaluations, including for clinical trials and personalized oncology approaches.

Micromanipulation of Live Microdissected Tissues with a Low-Cost Integrated Robotic Platform.

The small amount of human tissue available for testing is a paramount challenge in cancer drug development, cancer disease models, and personalized oncology. Technologies that combine the microscale manipulation of tissues with fluid handling offer the exciting possibility of miniaturizing and automating drug evaluation workflows. This approach minimizes animal testing and enables inexpensive, more efficient testing of samples with high clinical biomimicry using scarce materials. We have developed an inexpensive platform based on an off-the-shelf robot that can manipulate microdissected tissues (µDTs) into user-programmed positions without using intricate microfluidic designs nor any other accessories such as a microscope or a pneumatic controller. The robot integrates complex functions such as vision and fluid actuation by incorporating simple items including a USB camera and a rotary pump. Through the robot's camera, the platform software optically recognizes randomly-seeded µDTs on the surface of a petri dish and positions a mechanical arm above the µDTs. Then, a custom rotary pump actuated by one of the robot's motors generates enough microfluidic lift to hydrodynamically pick and place µDTs with a pipette at a safe distance from the substrate without requiring a proximity sensor. The platform's simple, integrated construction is cost-effective and compact, allowing placement inside a tissue culture hood for sterile workflows. The platform enables users to select µDTs based on their size, place them in user-programmed arrays, such as multi-well plates, and control various robot motion parameters. As a case application, we use the robotic system to conduct semi-automated drug testing of mouse and human µDTs in 384-well plates. Our user-friendly platform promises to democratize microscale tissue research to clinical and biological laboratories worldwide.

DNAJB1-PRKACA fusion protein-regulated LINC00473 promotes tumor growth and alters mitochondrial fitness in fibrolamellar carcinoma.

Fibrolamellar carcinoma (FLC) is a rare liver cancer that disproportionately affects adolescents and young adults. Currently, no standard of care is available and there remains a dire need for new therapeutics. Most patients harbor the fusion oncogene DNAJB1-PRKACA (DP fusion), but clinical inhibitors are not yet developed and it is critical to identify downstream mediators of FLC pathogenesis. Here, we identify long noncoding RNA LINC00473 among the most highly upregulated genes in FLC tumors and determine that it is strongly suppressed by RNAi-mediated inhibition of the DP fusion in FLC tumor epithelial cells. We show by loss- and gain-of-function studies that LINC00473 suppresses apoptosis, increases the expression of FLC marker genes, and promotes FLC growth in cell-based and in vivo disease models. Mechanistically, LINC00473 plays an important role in promoting glycolysis and altering mitochondrial activity. Specifically, LINC00473 knockdown leads to increased spare respiratory capacity, which indicates mitochondrial fitness. Overall, we propose that LINC00473 could be a viable target for this devastating disease.

A kinase to cytokine explorer to identify molecular regulators and potential therapeutic opportunities.

Cytokines and chemokines are secreted proteins that regulate various biological processes, such as inflammation, immune response, and cell differentiation. Therefore, disruption of signaling pathways involving these proteins has been linked to a range of diseases, including cancer. However, targeting individual cytokines, chemokines, or their receptors is challenging due to their regulatory redundancy and incomplete understanding of their signaling networks. To transform these difficult-to-drug targets into a pharmacologically manageable class, we developed a web-based platform called KinCytE. This platform was designed to link the effects of kinase inhibitors, a well-established class of drugs, with cytokine and chemokine release and signaling networks. The resulting KinCytE platform enables users to investigate protein kinases that regulate specific cytokines or chemokines, generate a ranked list of FDA-approved kinase inhibitors that affect cytokine/chemokine activity, and explore and visualize cytokine signaling network thus facilitating drugging this challenging target class. KinCytE is freely accessible via https://atlas.fredhutch.org/kincyte.

Exploiting the therapeutic vulnerability of IDH-mutant gliomas with zotiraciclib.

Isocitrate dehydrogenase (IDH)-mutant gliomas have distinctive metabolic and biological traits that may render them susceptible to targeted treatments. Here, by conducting a high-throughput drug screen, we pinpointed a specific susceptibility of IDH-mutant gliomas to zotiraciclib (ZTR). ZTR exhibited selective growth inhibition across multiple IDH-mutant glioma in vitro and in vivo models. Mechanistically, ZTR at low doses suppressed CDK9 and RNA Pol II phosphorylation in IDH-mutant cells, disrupting mitochondrial function and NAD+ production, causing oxidative stress. Integrated biochemical profiling of ZTR kinase targets and transcriptomics unveiled that ZTR-induced bioenergetic failure was linked to the suppression of PIM kinase activity. We posit that the combination of mitochondrial dysfunction and an inability to adapt to oxidative stress resulted in significant cell death upon ZTR treatment, ultimately increasing the therapeutic vulnerability of IDH-mutant gliomas. These findings prompted a clinical trial evaluating ZTR in IDH-mutant gliomas towards precision medicine ( NCT05588141 ). Highlights: Zotiraciclib (ZTR), a CDK9 inhibitor, hinders IDH-mutant glioma growth in vitro and in vivo . ZTR halts cell cycle, disrupts respiration, and induces oxidative stress in IDH-mutant cells.ZTR unexpectedly inhibits PIM kinases, impacting mitochondria and causing bioenergetic failure.These findings led to the clinical trial NCT05588141, evaluating ZTR for IDH-mutant gliomas.

Organotypic Models for Functional Drug Testing of Human Cancers.

In the era of personalized oncology, there have been accelerated efforts to develop clinically relevant platforms to test drug sensitivities of individual cancers. An ideal assay will serve as a diagnostic companion to inform the oncologist of the various treatments that are sensitive and insensitive, thus improving outcome while minimizing unnecessary toxicities and costs. To date, no such platform exists for clinical use, but promising approaches are on the horizon that take advantage of improved techniques in creating human cancer models that encompass the entire tumor microenvironment, alongside technologies for assessing and analyzing tumor response. This review summarizes a number of current strategies that make use of intact human cancer tissues as organotypic cultures in drug sensitivity testing.

Polypharmacology-based kinome screen identifies new regulators of KSHV reactivation.

Kaposi's sarcoma-associated herpesvirus (KSHV) causes several human diseases including Kaposi's sarcoma (KS), a leading cause of cancer in Africa and in patients with AIDS. KS tumor cells harbor KSHV predominantly in a latent form, while typically <5% contain lytic replicating virus. Because both latent and lytic stages likely contribute to cancer initiation and progression, continued dissection of host regulators of this biological switch will provide insights into fundamental pathways controlling the KSHV life cycle and related disease pathogenesis. Several cellular protein kinases have been reported to promote or restrict KSHV reactivation, but our knowledge of these signaling mediators and pathways is incomplete. We employed a polypharmacology-based kinome screen to identify specific kinases that regulate KSHV reactivation. Those identified by the screen and validated by knockdown experiments included several kinases that enhance lytic reactivation: ERBB2 (HER2 or neu), ERBB3 (HER3), ERBB4 (HER4), MKNK2 (MNK2), ITK, TEC, and DSTYK (RIPK5). Conversely, ERBB1 (EGFR1 or HER1), MKNK1 (MNK1) and FRK (PTK5) were found to promote the maintenance of latency. Mechanistic characterization of ERBB2 pro-lytic functions revealed a signaling connection between ERBB2 and the activation of CREB1, a transcription factor that drives KSHV lytic gene expression. These studies provided a proof-of-principle application of a polypharmacology-based kinome screen for the study of KSHV reactivation and enabled the discovery of both kinase inhibitors and specific kinases that regulate the KSHV latent-to-lytic replication switch.

Anesthetics inhibit phosphorylation of the ribosomal protein S6 in mouse cultured cortical cells and developing brain.

Introduction: The development and maintenance of neural circuits is highly sensitive to neural activity. General anesthetics have profound effects on neural activity and, as such, there is concern that these agents may alter cellular integrity and interfere with brain wiring, such as when exposure occurs during the vulnerable period of brain development. Under those conditions, exposure to anesthetics in clinical use today causes changes in synaptic strength and number, widespread apoptosis, and long-lasting cognitive impairment in a variety of animal models. Remarkably, most anesthetics produce these effects despite having differing receptor mechanisms of action. We hypothesized that anesthetic agents mediate these effects by inducing a shared signaling pathway. Methods: We exposed cultured cortical cells to propofol, etomidate, or dexmedetomidine and assessed the protein levels of dozens of signaling molecules and post-translational modifications using reverse phase protein arrays. To probe the role of neural activity, we performed separate control experiments to alter neural activity with non-anesthetics. Having identified anesthetic-induced changes in vitro, we investigated expression of the target proteins in the cortex of sevoflurane anesthetized postnatal day 7 mice by Western blotting. Results: All the anesthetic agents tested in vitro reduced phosphorylation of the ribosomal protein S6, an important member of the mTOR signaling pathway. We found a comparable decrease in cortical S6 phosphorylation by Western blotting in sevoflurane anesthetized neonatal mice. Using a systems approach, we determined that propofol, etomidate, dexmedetomidine, and APV/TTX all similarly modulate a signaling module that includes pS6 and other cell mediators of the mTOR-signaling pathway. Discussion: Reduction in S6 phosphorylation and subsequent suppression of the mTOR pathway may be a common and novel signaling event that mediates the impact of general anesthetics on neural circuit development.

Blockade of interleukin 10 potentiates antitumour immune function in human colorectal cancer liver metastases.

OBJECTIVE: Programmed cell death protein 1 (PD-1) checkpoint inhibition and adoptive cellular therapy have had limited success in patients with microsatellite stable colorectal cancer liver metastases (CRLM). We sought to evaluate the effect of interleukin 10 (IL-10) blockade on endogenous T cell and chimeric antigen receptor T (CAR-T) cell antitumour function in CRLM slice cultures. DESIGN: We created organotypic slice cultures from human CRLM (n=38 patients' tumours) and tested the antitumour effects of a neutralising antibody against IL-10 (αIL-10) both alone as treatment and in combination with exogenously administered carcinoembryonic antigen (CEA)-specific CAR-T cells. We evaluated slice cultures with single and multiplex immunohistochemistry, in situ hybridisation, single-cell RNA sequencing, reverse-phase protein arrays and time-lapse fluorescent microscopy. RESULTS: αIL-10 generated a 1.8-fold increase in T cell-mediated carcinoma cell death in human CRLM slice cultures. αIL-10 significantly increased proportions of CD8+ T cells without exhaustion transcription changes, and increased human leukocyte antigen - DR isotype (HLA-DR) expression of macrophages. The antitumour effects of αIL-10 were reversed by major histocompatibility complex class I or II (MHC-I or MHC-II) blockade, confirming the essential role of antigen presenting cells. Interrupting IL-10 signalling also rescued murine CAR-T cell proliferation and cytotoxicity from myeloid cell-mediated immunosuppression. In human CRLM slices, αIL-10 increased CEA-specific CAR-T cell activation and CAR-T cell-mediated cytotoxicity, with nearly 70% carcinoma cell apoptosis across multiple human tumours. Pretreatment with an IL-10 receptor blocking antibody also potentiated CAR-T function. CONCLUSION: Neutralising the effects of IL-10 in human CRLM has therapeutic potential as a stand-alone treatment and to augment the function of adoptively transferred CAR-T cells.

Subtype and Site Specific-Induced Metabolic Vulnerabilities in Prostate Cancer.

Aberrant metabolic functions play a crucial role in prostate cancer progression and lethality. Currently, limited knowledge is available on subtype-specific metabolic features and their implications for treatment. We therefore investigated the metabolic determinants of the two major subtypes of castration-resistant prostate cancer [androgen receptor-expressing prostate cancer (ARPC) and aggressive variant prostate cancer (AVPC)]. Transcriptomic analyses revealed enrichment of gene sets involved in oxidative phosphorylation (OXPHOS) in ARPC tumor samples compared with AVPC. Unbiased screening of metabolic signaling pathways in patient-derived xenograft models by proteomic analyses further supported an enrichment of OXPHOS in ARPC compared with AVPC, and a skewing toward glycolysis by AVPC. In vitro, ARPC C4-2B cells depended on aerobic respiration, while AVPC PC3 cells relied more heavily on glycolysis, as further confirmed by pharmacologic interference using IACS-10759, a clinical-grade inhibitor of OXPHOS. In vivo studies confirmed IACS-10759's inhibitory effects in subcutaneous and bone-localized C4-2B tumors, and no effect in subcutaneous PC3 tumors. Unexpectedly, IACS-10759 inhibited PC3 tumor growth in bone, indicating microenvironment-induced metabolic reprogramming. These results suggest that castration-resistant ARPC and AVPC exhibit different metabolic dependencies, which can further undergo metabolic reprogramming in bone. IMPLICATIONS: These vulnerabilities may be exploited with mechanistically novel treatments, such as those targeting OXPHOS alone or possibly in combination with existing therapies. In addition, our findings underscore the impact of the tumor microenvironment in reprogramming prostate cancer metabolism.

Both YAP1-MAML2 and constitutively active YAP1 drive the formation of tumors that resemble NF2 mutant meningiomas in mice.

YAP1 is a transcriptional coactivator regulated by the Hippo signaling pathway, including NF2. Meningiomas are the most common primary brain tumors; a large percentage exhibit heterozygous loss of chromosome 22 (harboring the NF2 gene) and functional inactivation of the remaining NF2 copy, implicating oncogenic YAP activity in these tumors. Recently, fusions between YAP1 and MAML2 have been identified in a subset of pediatric NF2 wild-type meningiomas. Here, we show that human YAP1-MAML2-positive meningiomas resemble NF2 mutant meningiomas by global and YAP-related gene expression signatures. We then show that expression of YAP1-MAML2 in mice induces tumors that resemble human YAP1 fusion-positive and NF2 mutant meningiomas by gene expression. We demonstrate that YAP1-MAML2 primarily functions by exerting TEAD-dependent YAP activity that is resistant to Hippo signaling. Treatment with YAP-TEAD inhibitors is sufficient to inhibit the viability of YAP1-MAML2-driven mouse tumors ex vivo. Finally, we show that expression of constitutively active YAP1 (S127/397A-YAP1) is sufficient to induce similar tumors, suggesting that the YAP component of the gene fusion is the critical driver of these tumors. In summary, our results implicate YAP1-MAML2 as a causal oncogenic driver and highlight TEAD-dependent YAP activity as an oncogenic driver in YAP1-MAML2 fusion meningioma as well as NF2 mutant meningioma in general.

Deep neural network modeling identifies biomarkers of response to immune-checkpoint therapy.

Immunotherapy has shown significant promise as a treatment for cancer, such as lung cancer and melanoma. However, only 10%-30% of the patients respond to treatment with immune checkpoint blockers (ICBs), underscoring the need for biomarkers to predict response to immunotherapy. Here, we developed DeepGeneX, a computational framework that uses advanced deep neural network modeling and feature elimination to reduce single-cell RNA-seq data on ∼26,000 genes to six of the most important genes (CCR7, SELL, GZMB, WARS, GZMH, and LGALS1), that accurately predict response to immunotherapy. We also discovered that the high LGALS1 and WARS-expressing macrophage population represent a biomarker for ICB therapy nonresponders, suggesting that these macrophages may be a target for improving ICB response. Taken together, DeepGeneX enables biomarker discovery and provides an understanding of the molecular basis for the model's predictions.

Olverembatinib inhibits SARS-CoV-2-Omicron variant-mediated cytokine release in human peripheral blood mononuclear cells.

The N-terminus domain (NTD) of the SARS-CoV-2 Omicron variant spike protien strongly induces multiple inflammatory molecules in human peripheral blood mononuclear cells, unaffected by the mutations observed in the NTD. Olverembatinib, a clinical-stage multi-kinase inhibitor, potently inhibits Omicron NTD-mediated cytokine release.

Proteomic and Transcriptomic Profiling Reveals Mitochondrial Oxidative Phosphorylation as Therapeutic Vulnerability in Androgen Receptor Pathway Active Prostate Tumors.

Metastatic prostate cancer (PC) is the second leading cause of cancer deaths in males and has limited therapeutic options. The lack of preclinical models for advanced stage PC represents one of the primary barriers in understanding the key genetic drivers of aggressive subsets, including androgen receptor (AR) pathway active and AR-null castration-resistant prostate cancers (CRPC). In our studies, we described a series of LuCaP patient-derived xenograft (PDX) models representing the major genomic and phenotypic features of human disease. To fully exploit the potential of these preclinical models, we carried out a comprehensive transcriptomic and proteomic profiling of 42 LuCaP PDX prostate tumors. The collected proteomic data (~6000 data points) based on 71 antibodies revealed many of the previously known molecular markers associated with AR-positive and AR-null CRPC. Genomic analysis indicated subtype-specific activation of pathways such as Wnt/beta-catenin signaling, mTOR, and oxidative phosphorylation for AR-positive CRPC and upregulation of carbohydrate metabolism and glucose metabolism for AR-null CRPC. Of these, we functionally confirmed the role of mitochondrial metabolism in AR-positive CRPC cell lines. Our data highlight how the integration of transcriptomic and proteomic approaches and PDX systems as preclinical models can potentially map the connectivity of poorly understood signaling pathways in metastatic prostate cancer.

Polypharmacologic Reprogramming of Tumor-Associated Macrophages toward an Inflammatory Phenotype.

Tumor-associated macrophages (TAM) are an important component of the tumor microenvironment (TME) that can promote tumor progression, metastasis, and resistance to therapies. Although TAMs represent a promising target for therapeutic intervention, the complexity of the TME has made the study of TAMs challenging. Here, we established a physiologically relevant in vitro TAM polarization system that recapitulates TAM protumoral activities. This system was used to characterize dynamic changes in gene expression and protein phosphorylation during TAM polarization and to screen phenotypic kinase inhibitors that impact TAM programming. BMS-794833, a multitargeted compound, was identified as a potent inhibitor of TAM polarization. BMS-794833 decreased protumoral properties of TAMs in vitro and suppressed tumor growth in mouse triple-negative breast cancer models. The effect of BMS-794833 was independent of its primary targets (MET and VEGFR2) but was dependent on its effect on multiple signaling pathways, including focal adhesion kinases, SRC family kinases, STAT3, and p38 MAPKs. Collectively, these findings underline the efficacy of polypharmacologic strategies in reprogramming complex signaling cascades activated during TAM polarization. SIGNIFICANCE: A physiologically relevant in vitro system of TAM polarization uncovers signaling pathways that regulate polarization and identifies strategies to target macrophage reprogramming to suppress cancer growth.

Utilizing preclinical models to develop targeted therapies for rare central nervous system cancers.

Patients with rare central nervous system (CNS) tumors typically have a poor prognosis and limited therapeutic options. Historically, these cancers have been difficult to study due to small number of patients. Recent technological advances have identified molecular drivers of some of these rare cancers which we can now use to generate representative preclinical models of these diseases. In this review, we outline the advantages and disadvantages of different models, emphasizing the utility of various in vitro and ex vivo models for target discovery and mechanistic inquiry and multiple in vivo models for therapeutic validation. We also highlight recent literature on preclinical model generation and screening approaches for ependymomas, histone mutated high-grade gliomas, and atypical teratoid rhabdoid tumors, all of which are rare CNS cancers that have recently established genetic or epigenetic drivers. These preclinical models are critical to advancing targeted therapeutics for these rare CNS cancers that currently rely on conventional treatments.

Computational modeling identifies multitargeted kinase inhibitors as effective therapies for metastatic, castration-resistant prostate cancer.

Castration-resistant prostate cancer (CRPC) is an advanced subtype of prostate cancer with limited therapeutic options. Here, we applied a systems-based modeling approach called kinome regularization (KiR) to identify multitargeted kinase inhibitors (KIs) that abrogate CRPC growth. Two predicted KIs, PP121 and SC-1, suppressed CRPC growth in two-dimensional in vitro experiments and in vivo subcutaneous xenografts. An ex vivo bone mimetic environment and in vivo tibia xenografts revealed resistance to these KIs in bone. Combining PP121 or SC-1 with docetaxel, standard-of-care chemotherapy for late-stage CRPC, significantly reduced tibia tumor growth in vivo, decreased growth factor signaling, and vastly extended overall survival, compared to either docetaxel monotherapy. These results highlight the utility of computational modeling in forming physiologically relevant predictions and provide evidence for the role of multitargeted KIs as chemosensitizers for late-stage, metastatic CRPC.

Machine learning identifies molecular regulators and therapeutics for targeting SARS-CoV2-induced cytokine release.

Although 15-20% of COVID-19 patients experience hyper-inflammation induced by massive cytokine production, cellular triggers of this process and strategies to target them remain poorly understood. Here, we show that the N-terminal domain (NTD) of the SARS-CoV-2 spike protein substantially induces multiple inflammatory molecules in myeloid cells and human PBMCs. Using a combination of phenotypic screening with machine learning-based modeling, we identified and experimentally validated several protein kinases, including JAK1, EPHA7, IRAK1, MAPK12, and MAP3K8, as essential downstream mediators of NTD-induced cytokine production, implicating the role of multiple signaling pathways in cytokine release. Further, we found several FDA-approved drugs, including ponatinib, and cobimetinib as potent inhibitors of the NTD-mediated cytokine release. Treatment with ponatinib outperforms other drugs, including dexamethasone and baricitinib, inhibiting all cytokines in response to the NTD from SARS-CoV-2 and emerging variants. Finally, ponatinib treatment inhibits lipopolysaccharide-mediated cytokine release in myeloid cells in vitro and lung inflammation mouse model. Together, we propose that agents targeting multiple kinases required for SARS-CoV-2-mediated cytokine release, such as ponatinib, may represent an attractive therapeutic option for treating moderate to severe COVID-19.

KiRNet: Kinase-centered network propagation of pharmacological screen results.

The ever-increasing size and scale of biological information have popularized network-based approaches as a means to interpret these data. We develop a network propagation method that integrates kinase-inhibitor-focused functional screens with known protein-protein interactions (PPIs). This method, dubbed KiRNet, uses an a priori edge-weighting strategy based on node degree to establish a pipeline from a kinase inhibitor screen to the generation of a predictive PPI subnetwork. We apply KiRNet to uncover molecular regulators of mesenchymal cancer cells driven by overexpression of Frizzled 2 (FZD2). KiRNet produces a network model consisting of 166 high-value proteins. These proteins exhibit FZD2-dependent differential phosphorylation, and genetic knockdown studies validate their role in maintaining a mesenchymal cell state. Finally, analysis of clinical data shows that mesenchymal tumors exhibit significantly higher average expression of the 166 corresponding genes than epithelial tumors for nine different cancer types.