The Rise of Hypothesis-Driven Artificial Intelligence in Oncology.

Cancer is a complex disease involving the deregulation of intricate cellular systems beyond genetic aberrations and, as such, requires sophisticated computational approaches and high-dimensional data for optimal interpretation. While conventional artificial intelligence (AI) models excel in many prediction tasks, they often lack interpretability and are blind to the scientific hypotheses generated by researchers to enable cancer discoveries. Here we propose that hypothesis-driven AI, a new emerging class of AI algorithm, is an innovative approach to uncovering the complex etiology of cancer from big omics data. This review exemplifies how hypothesis-driven AI is different from conventional AI by citing its application in various areas of oncology including tumor classification, patient stratification, cancer gene discovery, drug response prediction, and tumor spatial organization. Our aim is to stress the feasibility of incorporating domain knowledge and scientific hypotheses to craft the design of new AI algorithms. We showcase the power of hypothesis-driven AI in making novel cancer discoveries that can be overlooked by conventional AI methods. Since hypothesis-driven AI is still in its infancy, open questions such as how to better incorporate new knowledge and biological perspectives to ameliorate bias and improve interpretability in the design of AI algorithms still need to be addressed. In conclusion, hypothesis-driven AI holds great promise in the discovery of new mechanistic and functional insights that explain the complexity of cancer etiology and potentially chart a new roadmap to improve treatment regimens for individual patients.

MRI Manifestations of Breast Cancer Stroma and their Role in Predicting Molecular Subtype: A Case-control Study.

OBJECTIVE: This study explored whether breast MRI manifestations could be used to predict the stroma distribution of breast cancer (BC) and the role of tumor stroma-based MRI manifestations in molecular subtype prediction. METHODS: 57 patients with pathologically confirmed invasive BC (non-special type) who had lumpy BC on MRI within one week before surgery were retrospectively collected in the study. Stroma distributions were classified according to their characteristics in the pathological sections. The stromal distribution patterns among molecular subtypes were compared with the MRI manifestations of BC with different stroma distribution types (SDTs). RESULTS: SDTs were significantly different and depended on the BC hormone receptor (HR) (P<0.001). There were also significant differences among five SDTs on T2WI, ADC map, internal delayed enhanced features (IDEF), marginal delayed enhanced features (MDEF), and time signal intensity (TSI) curves. Spiculated margin and the absence of type-I TSI were independent predictors for BC with star grid type stroma. The appearance frequency of hypo-intensity on T2WI in HR- BCs was significantly lower (P=0.043) than in HR+ BCs. Star grid stroma and spiculated margin were key factors in predicting HR+ BCs, and the AUC was 0.927 (95% CI: 0.867-0.987). CONCLUSION: Breast MRI can be used to predict BC's stromal distribution and molecular subtypes.

Multiorgan locked-state model of chronic diseases and systems pharmacology opportunities.

With increasing human life expectancy, the global medical burden of chronic diseases is growing. Hence, chronic diseases are a pressing health concern and will continue to be in decades to come. Chronic diseases often involve multiple malfunctioning organs in the body. An imminent question is how interorgan crosstalk contributes to the etiology of chronic diseases. We conceived the locked-state model (LoSM), which illustrates how interorgan communication can give rise to body-wide memory-like properties that 'lock' healthy or pathological conditions. Next, we propose cutting-edge systems biology and artificial intelligence strategies to decipher chronic multiorgan locked states. Finally, we discuss the clinical implications of the LoSM and assess the power of systems-based therapies to dismantle pathological multiorgan locked states while improving treatments for chronic diseases.

SPIN-AI: A Deep Learning Model That Identifies Spatially Predictive Genes.

Spatially resolved sequencing technologies help us dissect how cells are organized in space. Several available computational approaches focus on the identification of spatially variable genes (SVGs), genes whose expression patterns vary in space. The detection of SVGs is analogous to the identification of differentially expressed genes and permits us to understand how genes and associated molecular processes are spatially distributed within cellular niches. However, the expression activities of SVGs fail to encode all information inherent in the spatial distribution of cells. Here, we devised a deep learning model, Spatially Informed Artificial Intelligence (SPIN-AI), to identify spatially predictive genes (SPGs), whose expression can predict how cells are organized in space. We used SPIN-AI on spatial transcriptomic data from squamous cell carcinoma (SCC) as a proof of concept. Our results demonstrate that SPGs not only recapitulate the biology of SCC but also identify genes distinct from SVGs. Moreover, we found a substantial number of ribosomal genes that were SPGs but not SVGs. Since SPGs possess the capability to predict spatial cellular organization, we reason that SPGs capture more biologically relevant information for a given cellular niche than SVGs. Thus, SPIN-AI has broad applications for detecting SPGs and uncovering which biological processes play important roles in governing cellular organization.

Network Biology-Inspired Machine Learning Features Predict Cancer Gene Targets and Reveal Target Coordinating Mechanisms.

Anticipating and understanding cancers' need for specific gene activities is key for novel therapeutic development. Here we utilized DepMap, a cancer gene dependency screen, to demonstrate that machine learning combined with network biology can produce robust algorithms that both predict what genes a cancer is dependent on and what network features coordinate such gene dependencies. Using network topology and biological annotations, we constructed four groups of novel engineered machine learning features that produced high accuracies when predicting binary gene dependencies. We found that in all examined cancer types, F1 scores were greater than 0.90, and model accuracy remained robust under multiple hyperparameter tests. We then deconstructed these models to identify tumor type-specific coordinators of gene dependency and identified that in certain cancers, such as thyroid and kidney, tumors' dependencies are highly predicted by gene connectivity. In contrast, other histologies relied on pathway-based features such as lung, where gene dependencies were highly predictive by associations with cell death pathway genes. In sum, we show that biologically informed network features can be a valuable and robust addition to predictive pharmacology models while simultaneously providing mechanistic insights.

Genetic predisposition to neuroblastoma results from a regulatory polymorphism that promotes the adrenergic cell state.

Childhood neuroblastomas exhibit plasticity between an undifferentiated neural crest-like mesenchymal cell state and a more differentiated sympathetic adrenergic cell state. These cell states are governed by autoregulatory transcriptional loops called core regulatory circuitries (CRCs), which drive the early development of sympathetic neuronal progenitors from migratory neural crest cells during embryogenesis. The adrenergic cell identity of neuroblastoma requires LMO1 as a transcriptional cofactor. Both LMO1 expression levels and the risk of developing neuroblastoma in children are associated with a single nucleotide polymorphism, G/T, that affects a GATA motif in the first intron of LMO1. Here, we showed that WT zebrafish with the GATA genotype developed adrenergic neuroblastoma, while knock-in of the protective TATA allele at this locus reduced the penetrance of MYCN-driven tumors, which were restricted to the mesenchymal cell state. Whole genome sequencing of childhood neuroblastomas demonstrated that TATA/TATA tumors also exhibited a mesenchymal cell state and were low risk at diagnosis. Thus, conversion of the regulatory GATA to a TATA allele in the first intron of LMO1 reduced the neuroblastoma-initiation rate by preventing formation of the adrenergic cell state. This mechanism was conserved over 400 million years of evolution, separating zebrafish and humans.

Manifold epigenetics: A conceptual model that guides engineering strategies to improve whole-body regenerative health.

Despite the promising advances in regenerative medicine, there is a critical need for improved therapies. For example, delaying aging and improving healthspan is an imminent societal challenge. Our ability to identify biological cues as well as communications between cells and organs are keys to enhance regenerative health and improve patient care. Epigenetics represents one of the major biological mechanisms involving in tissue regeneration, and therefore can be viewed as a systemic (body-wide) control. However, how epigenetic regulations concertedly lead to the development of biological memories at the whole-body level remains unclear. Here, we review the evolving definitions of epigenetics and identify missing links. We then propose our Manifold Epigenetic Model (MEMo) as a conceptual framework to explain how epigenetic memory arises and discuss what strategies can be applied to manipulate the body-wide memory. In summary we provide a conceptual roadmap for the development of new engineering approaches to improve regenerative health.

Genetic Predisposition to Neuroblastoma Results from a Regulatory Polymorphism that Promotes the Adrenergic Cell State.

Childhood neuroblastomas exhibit plasticity between an undifferentiated neural crest-like "mesenchymal" cell state and a more differentiated sympathetic "adrenergic" cell state. These cell states are governed by autoregulatory transcriptional loops called core regulatory circuitries (CRCs), which drive the early development of sympathetic neuronal progenitors from migratory neural crest cells during embryogenesis. The adrenergic cell identity of neuroblastoma requires LMO1 as a transcriptional co-factor. Both LMO1 expression levels and the risk of developing neuroblastoma in children are associated with a single nucleotide polymorphism G/T that affects a G ATA motif in the first intron of LMO1. Here we show that wild-type zebrafish with the G ATA genotype develop adrenergic neuroblastoma, while knock-in of the protective T ATA allele at this locus reduces the penetrance of MYCN-driven tumors, which are restricted to the mesenchymal cell state. Whole genome sequencing of childhood neuroblastomas demonstrates that T ATA/ T ATA tumors also exhibit a mesenchymal cell state and are low risk at diagnosis. Thus, conversion of the regulatory G ATA to a T ATA allele in the first intron of LMO1 reduces the neuroblastoma initiation rate by preventing formation of the adrenergic cell state, a mechanism that is conserved over 400 million years of evolution separating zebrafish and humans.

A Knowledge-Based Discovery Approach Couples Artificial Neural Networks With Weight Engineering to Uncover Immune-Related Processes Underpinning Clinical Traits of Breast Cancer.

Immune-related processes are important in underpinning the properties of clinical traits such as prognosis and drug response in cancer. The possibility to extract knowledge learned by artificial neural networks (ANNs) from omics data to explain cancer clinical traits is a very attractive subject for novel discovery. Recent studies using a version of ANNs called autoencoders revealed their capability to store biologically meaningful information indicating that autoencoders can be utilized as knowledge discovery platforms aside from their initial assigned use for dimensionality reduction. Here, we devise an innovative weight engineering approach and ANN platform called artificial neural network encoder (ANNE) using an autoencoder and apply it to a breast cancer dataset to extract knowledge learned by the autoencoder model that explains clinical traits. Intriguingly, the extracted biological knowledge in the form of gene-gene associations from ANNE shows immune-related components such as chemokines, carbonic anhydrase, and iron metabolism that modulate immune-related processes and the tumor microenvironment play important roles in underpinning breast cancer clinical traits. Our work shows that biological "knowledge" learned by an ANN model is indeed encoded as weights throughout its neuronal connections, and it is possible to extract learned knowledge via a novel weight engineering approach to uncover important biological insights.

Gene utility recapitulates chromosomal aberrancies in advanced stage neuroblastoma.

Neuroblastoma (NB) is the most common extracranial solid tumor in children. Although only a few recurrent somatic mutations have been identified, chromosomal abnormalities, including the loss of heterozygosity (LOH) at the chromosome 1p and gains of chromosome 17q, are often seen in the high-risk cases. The biological basis and evolutionary forces that drive such genetic abnormalities remain enigmatic. Here, we conceptualize the Gene Utility Model (GUM) that seeks to identify genes driving biological signaling via their collective gene utilities and apply it to understand the impact of those differentially utilized genes on constraining the evolution of NB karyotypes. By employing a computational process-guided flow algorithm to model gene utility in protein-protein networks that built based on transcriptomic data, we conducted several pairwise comparative analyses to uncover genes with differential utilities in stage 4 NBs with distinct classification. We then constructed a utility karyotype by mapping these differentially utilized genes to their respective chromosomal loci. Intriguingly, hotspots of the utility karyotype, to certain extent, can consistently recapitulate the major chromosomal abnormalities of NBs and also provides clues to yet identified predisposition sites. Hence, our study not only provides a new look, from a gene utility perspective, into the known chromosomal abnormalities detected by integrative genomic sequencing efforts, but also offers new insights into the etiology of NB and provides a framework to facilitate the identification of novel therapeutic targets for this devastating childhood cancer.

Metabolic Enzyme DLST Promotes Tumor Aggression and Reveals a Vulnerability to OXPHOS Inhibition in High-Risk Neuroblastoma.

High-risk neuroblastoma remains therapeutically challenging to treat, and the mechanisms promoting disease aggression are poorly understood. Here, we show that elevated expression of dihydrolipoamide S-succinyltransferase (DLST) predicts poor treatment outcome and aggressive disease in patients with neuroblastoma. DLST is an E2 component of the α-ketoglutarate (αKG) dehydrogenase complex, which governs the entry of glutamine into the tricarboxylic acid cycle (TCA) for oxidative decarboxylation. During this irreversible step, αKG is converted into succinyl-CoA, producing NADH for oxidative phosphorylation (OXPHOS). Utilizing a zebrafish model of MYCN-driven neuroblastoma, we demonstrate that even modest increases in DLST expression promote tumor aggression, while monoallelic dlst loss impedes disease initiation and progression. DLST depletion in human MYCN-amplified neuroblastoma cells minimally affected glutamine anaplerosis and did not alter TCA cycle metabolites other than αKG. However, DLST loss significantly suppressed NADH production and impaired OXPHOS, leading to growth arrest and apoptosis of neuroblastoma cells. In addition, multiple inhibitors targeting the electron transport chain, including the potent IACS-010759 that is currently in clinical testing for other cancers, efficiently reduced neuroblastoma proliferation in vitro. IACS-010759 also suppressed tumor growth in zebrafish and mouse xenograft models of high-risk neuroblastoma. Together, these results demonstrate that DLST promotes neuroblastoma aggression and unveils OXPHOS as an essential contributor to high-risk neuroblastoma. SIGNIFICANCE: These findings demonstrate a novel role for DLST in neuroblastoma aggression and identify the OXPHOS inhibitor IACS-010759 as a potential therapeutic strategy for this deadly disease.

Zebrafish as a Neuroblastoma Model: Progress Made, Promise for the Future.

For nearly a decade, researchers in the field of pediatric oncology have been using zebrafish as a model for understanding the contributions of genetic alternations to the pathogenesis of neuroblastoma (NB), and exploring the molecular and cellular mechanisms that underlie neuroblastoma initiation and metastasis. In this review, we will enumerate and illustrate the key advantages of using the zebrafish model in NB research, which allows researchers to: monitor tumor development in real-time; robustly manipulate gene expression (either transiently or stably); rapidly evaluate the cooperative interactions of multiple genetic alterations to disease pathogenesis; and provide a highly efficient and low-cost methodology to screen for effective pharmaceutical interventions (both alone and in combination with one another). This review will then list some of the common challenges of using the zebrafish model and provide strategies for overcoming these difficulties. We have also included visual diagram and figures to illustrate the workflow of cancer model development in zebrafish and provide a summary comparison of commonly used animal models in cancer research, as well as key findings of cooperative contributions between MYCN and diverse singling pathways in NB pathogenesis.

Zebrafish Model of Neuroblastoma Metastasis.

Zebrafish has emerged as an important animal model to study human diseases, especially cancer. Along with the robust transgenic and genome editing technologies applied in zebrafish modeling, the ease of maintenance, high-yield productivity, and powerful live imaging altogether make the zebrafish a valuable model system to study metastasis and cellular and molecular bases underlying this process in vivo. The first zebrafish neuroblastoma (NB) model of metastasis was developed by overexpressing two oncogenes, MYCN and LMO1, under control of the dopamine-beta-hydroxylase (dßh) promoter. Co-overexpressed MYCN and LMO1 led to the reduced latency and increased penetrance of neuroblastomagenesis, as well as accelerated distant metastasis of tumor cells. This new model reliably reiterates many key features of human metastatic NB, including involvement of clinically relevant and metastasis-associated genetic alterations; natural and spontaneous development of metastasis in vivo; and conserved sites of metastases. Therefore, the zebrafish model possesses unique advantages to dissect the complex process of tumor metastasis in vivo.

A novel combination therapy targeting ubiquitin-specific protease 5 in MYCN-driven neuroblastoma.

Histone deacetylase (HDAC) inhibitors are effective in MYCN-driven cancers, because of a unique need for HDAC recruitment by the MYCN oncogenic signal. However, HDAC inhibitors are much more effective in combination with other anti-cancer agents. To identify novel compounds which act synergistically with HDAC inhibitor, such as suberanoyl hydroxamic acid (SAHA), we performed a cell-based, high-throughput drug screen of 10,560 small molecule compounds from a drug-like diversity library and identified a small molecule compound (SE486-11) which synergistically enhanced the cytotoxic effects of SAHA. Effects of drug combinations on cell viability, proliferation, apoptosis and colony forming were assessed in a panel of neuroblastoma cell lines. Treatment with SAHA and SE486-11 increased MYCN ubiquitination and degradation, and markedly inhibited tumorigenesis in neuroblastoma xenografts, and, MYCN transgenic zebrafish and mice. The combination reduced ubiquitin-specific protease 5 (USP5) levels and increased unanchored polyubiquitin chains. Overexpression of USP5 rescued neuroblastoma cells from the cytopathic effects of the combination and reduced unanchored polyubiquitin, suggesting USP5 is a therapeutic target of the combination. SAHA and SE486-11 directly bound to USP5 and the drug combination exhibited a 100-fold higher binding to USP5 than individual drugs alone in microscale thermophoresis assays. MYCN bound to the USP5 promoter and induced USP5 gene expression suggesting that USP5 and MYCN expression created a forward positive feedback loop in neuroblastoma cells. Thus, USP5 acts as an oncogenic cofactor with MYCN in neuroblastoma and the novel combination of HDAC inhibitor with SE486-11 represents a novel therapeutic approach for the treatment of MYCN-driven neuroblastoma.

GAS7 Deficiency Promotes Metastasis in MYCN-Driven Neuroblastoma.

One of the greatest barriers to curative treatment of neuroblastoma is its frequent metastatic outgrowth prior to diagnosis, especially in cases driven by amplification of the MYCN oncogene. However, only a limited number of regulatory proteins that contribute to this complex MYCN-mediated process have been elucidated. Here we show that the growth arrest-specific 7 (GAS7) gene, located at chromosome band 17p13.1, is preferentially deleted in high-risk MYCN-driven neuroblastoma. GAS7 expression was also suppressed in MYCN-amplified neuroblastoma lacking 17p deletion. GAS7 deficiency led to accelerated metastasis in both zebrafish and mammalian models of neuroblastoma with overexpression or amplification of MYCN. Analysis of expression profiles and the ultrastructure of zebrafish neuroblastoma tumors with MYCN overexpression identified that GAS7 deficiency led to (i) downregulation of genes involved in cell-cell interaction, (ii) loss of contact among tumor cells as critical determinants of accelerated metastasis, and (iii) increased levels of MYCN protein. These results provide the first genetic evidence that GAS7 depletion is a critical early step in the cascade of events culminating in neuroblastoma metastasis in the context of MYCN overexpression. SIGNIFICANCE: Heterozygous deletion or MYCN-mediated repression of GAS7 in neuroblastoma releases an important brake on tumor cell dispersion and migration to distant sites, providing a novel mechanism underlying tumor metastasis in MYCN-driven neuroblastoma.See related commentary by Menard, p. 2815.

Ablation Behavior of a Carbon Fabric Reinforced Phenolic Composite Modified by Surface-Decorated ZrB2/SiC.

Carbon fabric reinforced phenolic composites were widely used as TPSs (thermal protection system) material in the aerospace industry. However, their limited oxidative ablation resistance restricted their further utility in more serious service conditions. In this study, the surface-decorated ZrB2/SiC and its modified carbon fabric reinforced phenolic composites have been successfully prepared. The self-modification mechanism of the surface-decorated ZrB2/SiC particles were characterized. The mechanical performance and ablation behavior of the composites were investigated. Results showed that the ZrB2/SiC particles possessed a good surface-decorated effect, which achieved good compatibility with the phenolic resin. The mechanical performance of the modified phenolic composite was effectively improved. The anti-oxidative ablation performance of the composite was improved. The mass ablation rate of the surface-decorated ZrB2-SiC-modified carbon fabric reinforced phenolic composites was 25% lower than that of the unmodified composites. The formed ZrO2 ceramic layer attached to the surface of the residual chars prevented the heat energy and oxygen from the inner material. Meanwhile, the volatilization of SiO2 and B2O3 effectively increased the heat dissipation. All these results confirmed that the ZrB2-SiC particles can effectively improve the ablation resistance of the composite, which provided a basis for the application of the composites to more serious service environments.

Drugging MYCN Oncogenic Signaling through the MYCN-PA2G4 Binding Interface.

MYCN is a major driver for the childhood cancer, neuroblastoma, however, there are no inhibitors of this target. Enhanced MYCN protein stability is a key component of MYCN oncogenesis and is maintained by multiple feedforward expression loops involving MYCN transactivation target genes. Here, we reveal the oncogenic role of a novel MYCN target and binding protein, proliferation-associated 2AG4 (PA2G4). Chromatin immunoprecipitation studies demonstrated that MYCN occupies the PA2G4 gene promoter, stimulating transcription. Direct binding of PA2G4 to MYCN protein blocked proteolysis of MYCN and enhanced colony formation in a MYCN-dependent manner. Using molecular modeling, surface plasmon resonance, and mutagenesis studies, we mapped the MYCN-PA2G4 interaction site to a 14 amino acid MYCN sequence and a surface crevice of PA2G4. Competitive chemical inhibition of the MYCN-PA2G4 protein-protein interface had potent inhibitory effects on neuroblastoma tumorigenesis in vivo. Treated tumors showed reduced levels of both MYCN and PA2G4. Our findings demonstrate a critical role for PA2G4 as a cofactor in MYCN-driven neuroblastoma and highlight competitive inhibition of the PA2G4-MYCN protein binding as a novel therapeutic strategy in the disease. SIGNIFICANCE: Competitive chemical inhibition of the PA2G4-MYCN protein interface provides a basis for drug design of small molecules targeting MYC and MYCN-binding partners in malignancies driven by MYC family oncoproteins.

MYC Drives a Subset of High-Risk Pediatric Neuroblastomas and Is Activated through Mechanisms Including Enhancer Hijacking and Focal Enhancer Amplification.

The amplified MYCN gene serves as an oncogenic driver in approximately 20% of high-risk pediatric neuroblastomas. Here, we show that the family member MYC is a potent transforming gene in a separate subset of high-risk neuroblastoma cases (∼10%), based on (i) its upregulation by focal enhancer amplification or genomic rearrangements leading to enhancer hijacking, and (ii) its ability to transform neuroblastoma precursor cells in a transgenic animal model. The aberrant regulatory elements associated with oncogenic MYC activation include focally amplified distal enhancers and translocation of highly active enhancers from other genes to within topologically associating domains containing the MYC gene locus. The clinical outcome for patients with high levels of MYC expression is virtually identical to that of patients with amplification of the MYCN gene, a known high-risk feature of this disease. Together, these findings establish MYC as a bona fide oncogene in a clinically significant group of high-risk childhood neuroblastomas.Significance: Amplification of the MYCN oncogene is a recognized hallmark of high-risk pediatric neuroblastoma. Here, we demonstrate that MYC is also activated as a potent oncogene in a distinct subset of neuroblastoma cases through either focal amplification of distal enhancers or enhancer hijacking mediated by chromosomal translocation. Cancer Discov; 8(3); 320-35. ©2017 AACR.This article is highlighted in the In This Issue feature, p. 253.

LMO1 Synergizes with MYCN to Promote Neuroblastoma Initiation and Metastasis.

A genome-wide association study identified LMO1, which encodes an LIM-domain-only transcriptional cofactor, as a neuroblastoma susceptibility gene that functions as an oncogene in high-risk neuroblastoma. Here we show that dßh promoter-mediated expression of LMO1 in zebrafish synergizes with MYCN to increase the proliferation of hyperplastic sympathoadrenal precursor cells, leading to a reduced latency and increased penetrance of neuroblastomagenesis. The transgenic expression of LMO1 also promoted hematogenous dissemination and distant metastasis, which was linked to neuroblastoma cell invasion and migration, and elevated expression levels of genes affecting tumor cell-extracellular matrix interaction, including loxl3, itga2b, itga3, and itga5. Our results provide in vivo validation of LMO1 as an important oncogene that promotes neuroblastoma initiation, progression, and widespread metastatic dissemination.