Next-generation characterization of the Cancer Cell Line Encyclopedia.

Large panels of comprehensively characterized human cancer models, including the Cancer Cell Line Encyclopedia (CCLE), have provided a rigorous framework with which to study genetic variants, candidate targets, and small-molecule and biological therapeutics and to identify new marker-driven cancer dependencies. To improve our understanding of the molecular features that contribute to cancer phenotypes, including drug responses, here we have expanded the characterizations of cancer cell lines to include genetic, RNA splicing, DNA methylation, histone H3 modification, microRNA expression and reverse-phase protein array data for 1,072 cell lines from individuals of various lineages and ethnicities. Integration of these data with functional characterizations such as drug-sensitivity, short hairpin RNA knockdown and CRISPR-Cas9 knockout data reveals potential targets for cancer drugs and associated biomarkers. Together, this dataset and an accompanying public data portal provide a resource for the acceleration of cancer research using model cancer cell lines.

CANDI: an R package and Shiny app for annotating radiographs and evaluating computer-aided diagnosis.

MOTIVATION: Radiologists have used algorithms for Computer-Aided Diagnosis (CAD) for decades. These algorithms use machine learning with engineered features, and there have been mixed findings on whether they improve radiologists' interpretations. Deep learning offers superior performance but requires more training data and has not been evaluated in joint algorithm-radiologist decision systems. RESULTS: We developed the Computer-Aided Note and Diagnosis Interface (CANDI) for collaboratively annotating radiographs and evaluating how algorithms alter human interpretation. The annotation app collects classification, segmentation, and image captioning training data, and the evaluation app randomizes the availability of CAD tools to facilitate clinical trials on radiologist enhancement. AVAILABILITY AND IMPLEMENTATION: Demonstrations and source code are hosted at (https://candi.nextgenhealthcare.org), and (https://github.com/mbadge/candi), respectively, under GPL-3 license. SUPPLEMENTARY INFORMATION: Supplementary material is available at Bioinformatics online.

Resistance mechanisms to TP53-MDM2 inhibition identified by in vivo piggyBac transposon mutagenesis screen in an Arf-/- mouse model.

Inhibitors of double minute 2 protein (MDM2)-tumor protein 53 (TP53) interaction are predicted to be effective in tumors in which the TP53 gene is wild type, by preventing TP53 protein degradation. One such setting is represented by the frequent CDKN2A deletion in human cancer that, through inactivation of p14ARF, activates MDM2 protein, which in turn degrades TP53 tumor suppressor. Here we used piggyBac (PB) transposon insertional mutagenesis to anticipate resistance mechanisms occurring during treatment with the MDM2-TP53 inhibitor HDM201. Constitutive PB mutagenesis in Arf-/- mice provided a collection of spontaneous tumors with characterized insertional genetic landscapes. Tumors were allografted in large cohorts of mice to assess the pharmacologic effects of HDM201. Sixteen out of 21 allograft models were sensitive to HDM201 but ultimately relapsed under treatment. A comparison of tumors with acquired resistance to HDM201 and untreated tumors identified 87 genes that were differentially and significantly targeted by the PB transposon. Resistant tumors displayed a complex clonality pattern suggesting the emergence of several resistant subclones. Among the most frequent alterations conferring resistance, we observed somatic and insertional loss-of-function mutations in transformation-related protein 53 (Trp53) in 54% of tumors and transposon-mediated gain-of-function alterations in B-cell lymphoma-extra large (Bcl-xL), Mdm4, and two TP53 family members, resulting in expression of the TP53 dominant negative truncations ΔNTrp63 and ΔNTrp73. Enhanced BCL-xL and MDM4 protein expression was confirmed in resistant tumors, as well as in HDM201-resistant patient-derived tumor xenografts. Interestingly, concomitant inhibition of MDM2 and BCL-xL demonstrated significant synergy in p53 wild-type cell lines in vitro. Collectively, our findings identify several potential mechanisms by which TP53 wild-type tumors may escape MDM2-targeted therapy.

Natural Language-based Machine Learning Models for the Annotation of Clinical Radiology Reports.

Purpose To compare different methods for generating features from radiology reports and to develop a method to automatically identify findings in these reports. Materials and Methods In this study, 96 303 head computed tomography (CT) reports were obtained. The linguistic complexity of these reports was compared with that of alternative corpora. Head CT reports were preprocessed, and machine-analyzable features were constructed by using bag-of-words (BOW), word embedding, and Latent Dirichlet allocation-based approaches. Ultimately, 1004 head CT reports were manually labeled for findings of interest by physicians, and a subset of these were deemed critical findings. Lasso logistic regression was used to train models for physician-assigned labels on 602 of 1004 head CT reports (60%) using the constructed features, and the performance of these models was validated on a held-out 402 of 1004 reports (40%). Models were scored by area under the receiver operating characteristic curve (AUC), and aggregate AUC statistics were reported for (a) all labels, (b) critical labels, and (c) the presence of any critical finding in a report. Sensitivity, specificity, accuracy, and F1 score were reported for the best performing model's (a) predictions of all labels and (b) identification of reports containing critical findings. Results The best-performing model (BOW with unigrams, bigrams, and trigrams plus average word embeddings vector) had a held-out AUC of 0.966 for identifying the presence of any critical head CT finding and an average 0.957 AUC across all head CT findings. Sensitivity and specificity for identifying the presence of any critical finding were 92.59% (175 of 189) and 89.67% (191 of 213), respectively. Average sensitivity and specificity across all findings were 90.25% (1898 of 2103) and 91.72% (18 351 of 20 007), respectively. Simpler BOW methods achieved results competitive with those of more sophisticated approaches, with an average AUC for presence of any critical finding of 0.951 for unigram BOW versus 0.966 for the best-performing model. The Yule I of the head CT corpus was 34, markedly lower than that of the Reuters corpus (at 103) or I2B2 discharge summaries (at 271), indicating lower linguistic complexity. Conclusion Automated methods can be used to identify findings in radiology reports. The success of this approach benefits from the standardized language of these reports. With this method, a large labeled corpus can be generated for applications such as deep learning. © RSNA, 2018 Online supplemental material is available for this article.

The Cancer Cell Line Encyclopedia enables predictive modelling of anticancer drug sensitivity.

The systematic translation of cancer genomic data into knowledge of tumour biology and therapeutic possibilities remains challenging. Such efforts should be greatly aided by robust preclinical model systems that reflect the genomic diversity of human cancers and for which detailed genetic and pharmacological annotation is available. Here we describe the Cancer Cell Line Encyclopedia (CCLE): a compilation of gene expression, chromosomal copy number and massively parallel sequencing data from 947 human cancer cell lines. When coupled with pharmacological profiles for 24 anticancer drugs across 479 of the cell lines, this collection allowed identification of genetic, lineage, and gene-expression-based predictors of drug sensitivity. In addition to known predictors, we found that plasma cell lineage correlated with sensitivity to IGF1 receptor inhibitors; AHR expression was associated with MEK inhibitor efficacy in NRAS-mutant lines; and SLFN11 expression predicted sensitivity to topoisomerase inhibitors. Together, our results indicate that large, annotated cell-line collections may help to enable preclinical stratification schemata for anticancer agents. The generation of genetic predictions of drug response in the preclinical setting and their incorporation into cancer clinical trial design could speed the emergence of 'personalized' therapeutic regimens.

Knocking out multigene redundancies via cycles of sexual assortment and fluorescence selection.

Phenotypes that might otherwise reveal a gene's function can be obscured by genes with overlapping function. This phenomenon is best known within gene families, in which an important shared function may only be revealed by mutating all family members. Here we describe the 'green monster' technology that enables precise deletion of many genes. In this method, a population of deletion strains with each deletion marked by an inducible green fluorescent protein reporter gene, is subjected to repeated rounds of mating, meiosis and flow-cytometric enrichment. This results in the aggregation of multiple deletion loci in single cells. The green monster strategy is potentially applicable to assembling other engineered alterations in any species with sex or alternative means of allelic assortment. To test the technology, we generated a single broadly drug-sensitive strain of Saccharomyces cerevisiae bearing precise deletions of all 16 ATP-binding cassette transporters within clades associated with multidrug resistance.

Using a wearable patch to develop a digital monitoring biomarker of inflammation in response to LPS challenge.

Remote inflammation monitoring with digital health technologies (DHTs) would provide valuable information for both clinical research and care. Controlled perturbations of the immune system may reveal physiological signatures which could be used to develop a digital biomarker of inflammatory state. In this study, molecular and physiological profiling was performed following an in vivo lipopolysaccharide (LPS) challenge to develop a digital biomarker of inflammation. Ten healthy volunteers received an intravenous LPS challenge and were monitored for 24 h using the VitalConnect VitalPatch (VitalPatch). VitalPatch measurements included heart rate (HR), heart rate variability (HRV), respiratory rate (RR), and skin temperature (TEMP). Conventional episodic inpatient vital signs and serum proteins were measured pre- and post-LPS challenge. The VitalPatch provided vital signs that were comparable to conventional methods for assessing HR, RR, and TEMP. A pronounced increase was observed in HR, RR, and TEMP as well as a decrease in HRV 1-4 h post-LPS challenge. The ordering of participants by magnitude of inflammatory cytokine response 2 h post-LPS challenge was consistent with ordering of participants by change from baseline in vital signs when measured by VitalPatch (r = 0.73) but not when measured by conventional methods (r = -0.04). A machine learning model trained on VitalPatch data predicted change from baseline in inflammatory protein response (R2 = 0.67). DHTs, such as VitalPatch, can improve upon existing episodic measurements of vital signs by enabling continuous sensing and have the potential for future use as tools to remotely monitor inflammation.

Current Status and Future Directions: The Application of Artificial Intelligence/Machine Learning for Precision Medicine.

Technological innovations, such as artificial intelligence (AI) and machine learning (ML), have the potential to expedite the goal of precision medicine, especially when combined with increased capacity for voluminous data from multiple sources and expanded therapeutic modalities; however, they also present several challenges. In this communication, we first discuss the goals of precision medicine, and contextualize the use of AI in precision medicine by showcasing innovative applications (e.g., prediction of tumor growth and overall survival, biomarker identification using biomedical images, and identification of patient population for clinical practice) which were presented during the February 2023 virtual public workshop entitled "Application of Artificial Intelligence and Machine Learning for Precision Medicine," hosted by the US Food and Drug Administration (FDA) and University of Maryland Center of Excellence in Regulatory Science and Innovation (M-CERSI). Next, we put forward challenges brought about by the multidisciplinary nature of AI, particularly highlighting the need for AI to be trustworthy. To address such challenges, we subsequently note practical approaches, viz., differential privacy, synthetic data generation, and federated learning. The proposed strategies - some of which are highlighted presentations from the workshop - are for the protection of personal information and intellectual property. In addition, methods such as the risk-based management approach and the need for an agile regulatory ecosystem are discussed. Finally, we lay out a call for action that includes sharing of data and algorithms, development of regulatory guidance documents, and pooling of expertise from a broad-spectrum of stakeholders to enhance the application of AI in precision medicine.

PGC-1alpha-responsive genes involved in oxidative phosphorylation are coordinately downregulated in human diabetes.

DNA microarrays can be used to identify gene expression changes characteristic of human disease. This is challenging, however, when relevant differences are subtle at the level of individual genes. We introduce an analytical strategy, Gene Set Enrichment Analysis, designed to detect modest but coordinate changes in the expression of groups of functionally related genes. Using this approach, we identify a set of genes involved in oxidative phosphorylation whose expression is coordinately decreased in human diabetic muscle. Expression of these genes is high at sites of insulin-mediated glucose disposal, activated by PGC-1alpha and correlated with total-body aerobic capacity. Our results associate this gene set with clinically important variation in human metabolism and illustrate the value of pathway relationships in the analysis of genomic profiling experiments.

Chemical combinations elucidate pathway interactions and regulation relevant to Hepatitis C replication.

The search for effective Hepatitis C antiviral therapies has recently focused on host sterol metabolism and protein prenylation pathways that indirectly affect viral replication. However, inhibition of the sterol pathway with statin drugs has not yielded consistent results in patients. Here, we present a combination chemical genetic study to explore how the sterol and protein prenylation pathways work together to affect hepatitis C viral replication in a replicon assay. In addition to finding novel targets affecting viral replication, our data suggest that the viral replication is strongly affected by sterol pathway regulation. There is a marked transition from antagonistic to synergistic antiviral effects as the combination targets shift downstream along the sterol pathway. We also show how pathway regulation frustrates potential hepatitis C therapies based on the sterol pathway, and reveal novel synergies that selectively inhibit hepatitis C replication over host toxicity. In particular, combinations targeting the downstream sterol pathway enzymes produced robust and selective synergistic inhibition of hepatitis C replication. Our findings show how combination chemical genetics can reveal critical pathway connections relevant to viral replication, and can identify potential treatments with an increased therapeutic window.

Combination chemical genetics.

Predicting the behavior of living organisms is an enormous challenge given their vast complexity. Efforts to model biological systems require large datasets generated by physical binding experiments and perturbation studies. Genetic perturbations have proven important and are greatly facilitated by the advent of comprehensive mutant libraries in model organisms. Small-molecule chemical perturbagens provide a complementary approach, especially for systems that lack mutant libraries, and can easily probe the function of essential genes. Though single chemical or genetic perturbations provide crucial information associating individual components (for example, genes, proteins or small molecules) with pathways or phenotypes, functional relationships between pathways and modules of components are most effectively obtained from combined perturbation experiments. Here we review the current state of and discuss some future directions for 'combination chemical genetics', the systematic application of multiple chemical or mixed chemical and genetic perturbations, both to gain insight into biological systems and to facilitate medical discoveries.

High-order combination effects and biological robustness.

Biological systems are robust, in that they can maintain stable phenotypes under varying conditions or attacks. Biological systems are also complex, being organized into many functional modules that communicate through interlocking pathways and feedback mechanisms. In these systems, robustness and complexity are linked because both qualities arise from the same underlying mechanisms. When perturbed by multiple attacks, such complex systems become fragile in both theoretical and experimental studies, and this fragility depends on the number of agents applied. We explore how this relationship can be used to study the functional robustness of a biological system using systematic high-order combination experiments. This presents a promising approach toward many biomedical and bioengineering challenges. For example, high-order experiments could determine the point of fragility for pathogenic bacteria and might help identify optimal treatments against multi-drug resistance. Such studies would also reinforce the growing appreciation that biological systems are best manipulated not by targeting a single protein, but by modulating the set of many nodes that can selectively control a system's functional state.

Quantifying neurologic disease using biosensor measurements in-clinic and in free-living settings in multiple sclerosis.

Technological advances in passive digital phenotyping present the opportunity to quantify neurological diseases using new approaches that may complement clinical assessments. Here, we studied multiple sclerosis (MS) as a model neurological disease for investigating physiometric and environmental signals. The objective of this study was to assess the feasibility and correlation of wearable biosensors with traditional clinical measures of disability both in clinic and in free-living in MS patients. This is a single site observational cohort study conducted at an academic neurological center specializing in MS. A cohort of 25 MS patients with varying disability scores were recruited. Patients were monitored in clinic while wearing biosensors at nine body locations at three separate visits. Biosensor-derived features including aspects of gait (stance time, turn angle, mean turn velocity) and balance were collected, along with standardized disability scores assessed by a neurologist. Participants also wore up to three sensors on the wrist, ankle, and sternum for 8 weeks as they went about their daily lives. The primary outcomes were feasibility, adherence, as well as correlation of biosensor-derived metrics with traditional neurologist-assessed clinical measures of disability. We used machine-learning algorithms to extract multiple features of motion and dexterity and correlated these measures with more traditional measures of neurological disability, including the expanded disability status scale (EDSS) and the MS functional composite-4 (MSFC-4). In free-living, sleep measures were additionally collected. Twenty-three subjects completed the first two of three in-clinic study visits and the 8-week free-living biosensor period. Several biosensor-derived features significantly correlated with EDSS and MSFC-4 scores derived at visit two, including mobility stance time with MSFC-4 z-score (Spearman correlation -0.546; p = 0.0070), several aspects of turning including turn angle (0.437; p = 0.0372), and maximum angular velocity (0.653; p = 0.0007). Similar correlations were observed at subsequent clinic visits, and in the free-living setting. We also found other passively collected signals, including measures of sleep, that correlated with disease severity. These findings demonstrate the feasibility of applying passive biosensor measurement techniques to monitor disability in MS patients both in clinic and in the free-living setting.

Chemical combination effects predict connectivity in biological systems.

Efforts to construct therapeutically useful models of biological systems require large and diverse sets of data on functional connections between their components. Here we show that cellular responses to combinations of chemicals reveal how their biological targets are connected. Simulations of pathways with pairs of inhibitors at varying doses predict distinct response surface shapes that are reproduced in a yeast experiment, with further support from a larger screen using human tumour cells. The response morphology yields detailed connectivity constraints between nearby targets, and synergy profiles across many combinations show relatedness between targets in the whole network. Constraints from chemical combinations complement genetic studies, because they probe different cellular components and can be applied to disease models that are not amenable to mutagenesis. Chemical probes also offer increased flexibility, as they can be continuously dosed, temporally controlled, and readily combined. After extending this initial study to cover a wider range of combination effects and pathway topologies, chemical combinations may be used to refine network models or to identify novel targets. This response surface methodology may even apply to non-biological systems where responses to targeted perturbations can be measured.

Automated deep-neural-network surveillance of cranial images for acute neurologic events.

Rapid diagnosis and treatment of acute neurological illnesses such as stroke, hemorrhage, and hydrocephalus are critical to achieving positive outcomes and preserving neurologic function-'time is brain'1-5. Although these disorders are often recognizable by their symptoms, the critical means of their diagnosis is rapid imaging6-10. Computer-aided surveillance of acute neurologic events in cranial imaging has the potential to triage radiology workflow, thus decreasing time to treatment and improving outcomes. Substantial clinical work has focused on computer-assisted diagnosis (CAD), whereas technical work in volumetric image analysis has focused primarily on segmentation. 3D convolutional neural networks (3D-CNNs) have primarily been used for supervised classification on 3D modeling and light detection and ranging (LiDAR) data11-15. Here, we demonstrate a 3D-CNN architecture that performs weakly supervised classification to screen head CT images for acute neurologic events. Features were automatically learned from a clinical radiology dataset comprising 37,236 head CTs and were annotated with a semisupervised natural-language processing (NLP) framework16. We demonstrate the effectiveness of our approach to triage radiology workflow and accelerate the time to diagnosis from minutes to seconds through a randomized, double-blinded, prospective trial in a simulated clinical environment.

Multi-target therapeutics: when the whole is greater than the sum of the parts.

Drugs designed to act against individual molecular targets cannot usually combat multigenic diseases such as cancer, or diseases that affect multiple tissues or cell types such as diabetes and immunoinflammatory disorders. Combination drugs that impact multiple targets simultaneously are better at controlling complex disease systems, are less prone to drug resistance and are the standard of care in many important therapeutic areas. The combination drugs currently employed are primarily of rational design, but the increased efficacy they provide justifies in vitro discovery efforts for identifying novel multi-target mechanisms. In this review, we discuss the biological rationale for combination therapeutics, review some existing combination drugs and present a systematic approach to identify interactions between molecular pathways that could be leveraged for therapeutic benefit.

Simulating Serial-Target Antibacterial Drug Synergies Using Flux Balance Analysis.

Flux balance analysis (FBA) is an increasingly useful approach for modeling the behavior of metabolic systems. However, standard FBA modeling of genetic knockouts cannot predict drug combination synergies observed between serial metabolic targets, even though such synergies give rise to some of the most widely used antibiotic treatments. Here we extend FBA modeling to simulate responses to chemical inhibitors at varying concentrations, by diverting enzymatic flux to a waste reaction. This flux diversion yields very similar qualitative predictions to prior methods for single target activity. However, we find very different predictions for combinations, where flux diversion, which mimics the kinetics of competitive metabolic inhibitors, can explain serial target synergies between metabolic enzyme inhibitors that we confirmed in Escherichia coli cultures. FBA flux diversion opens the possibility for more accurate genome-scale predictions of drug synergies, which can be used to suggest treatments for infections and other diseases.

High-Order Drug Combinations Are Required to Effectively Kill Colorectal Cancer Cells.

Like classical chemotherapy regimens used to treat cancer, targeted therapies will also rely upon polypharmacology, but tools are still lacking to predict which combinations of molecularly targeted drugs may be most efficacious. In this study, we used image-based proliferation and apoptosis assays in colorectal cancer cell lines to systematically investigate the efficacy of combinations of two to six drugs that target critical oncogenic pathways. Drug pairs targeting key signaling pathways resulted in synergies across a broad spectrum of genetic backgrounds but often yielded only cytostatic responses. Enhanced cytotoxicity was observed when additional processes including apoptosis and cell cycle were targeted as part of the combination. In some cases, where cell lines were resistant to paired and tripled drugs, increased expression of antiapoptotic proteins was observed, requiring a fourth-order combination to induce cytotoxicity. Our results illustrate how high-order drug combinations are needed to kill drug-resistant cancer cells, and they also show how systematic drug combination screening together with a molecular understanding of drug responses may help define optimal cocktails to overcome aggressive cancers. Cancer Res; 76(23); 6950-63. ©2016 AACR.

Gene Expression Ratios Lead to Accurate and Translatable Predictors of DR5 Agonism across Multiple Tumor Lineages.

Death Receptor 5 (DR5) agonists demonstrate anti-tumor activity in preclinical models but have yet to demonstrate robust clinical responses. A key limitation may be the lack of patient selection strategies to identify those most likely to respond to treatment. To overcome this limitation, we screened a DR5 agonist Nanobody across >600 cell lines representing 21 tumor lineages and assessed molecular features associated with response. High expression of DR5 and Casp8 were significantly associated with sensitivity, but their expression thresholds were difficult to translate due to low dynamic ranges. To address the translational challenge of establishing thresholds of gene expression, we developed a classifier based on ratios of genes that predicted response across lineages. The ratio classifier outperformed the DR5+Casp8 classifier, as well as standard approaches for feature selection and classification using genes, instead of ratios. This classifier was independently validated using 11 primary patient-derived pancreatic xenograft models showing perfect predictions as well as a striking linearity between prediction probability and anti-tumor response. A network analysis of the genes in the ratio classifier captured important biological relationships mediating drug response, specifically identifying key positive and negative regulators of DR5 mediated apoptosis, including DR5, CASP8, BID, cFLIP, XIAP and PEA15. Importantly, the ratio classifier shows translatability across gene expression platforms (from Affymetrix microarrays to RNA-seq) and across model systems (in vitro to in vivo). Our approach of using gene expression ratios presents a robust and novel method for constructing translatable biomarkers of compound response, which can also probe the underlying biology of treatment response.