Collaborative artificial intelligence system for investigation of healthcare claims compliance.

Healthcare fraud, waste and abuse are costly problems that have huge impact on society. Traditional approaches to identify non-compliant claims rely on auditing strategies requiring trained professionals, or on machine learning methods requiring labelled data and possibly lacking interpretability. We present Clais, a collaborative artificial intelligence system for claims analysis. Clais automatically extracts human-interpretable rules from healthcare policy documents (0.72 F1-score), and it enables professionals to edit and validate the extracted rules through an intuitive user interface. Clais executes the rules on claim records to identify non-compliance: on this task Clais significantly outperforms two baseline machine learning models, and its median F1-score is 1.0 (IQR = 0.83 to 1.0) when executing the extracted rules, and 1.0 (IQR = 1.0 to 1.0) when executing the same rules after human curation. Professionals confirm through a user study the usefulness of Clais in making their workflow simpler and more effective.

Novel WRN Helicase Inhibitors Selectively Target Microsatellite Unstable Cancer Cells.

Microsatellite-unstable (MSI) cancers require WRN helicase to resolve replication stress due to expanded DNA (TA)n-dinucleotide repeats. WRN is a promising synthetic lethal target for MSI tumours, and WRN inhibitors are in development. Here, we used CRISPR-Cas9 base editing to map WRN residues critical for MSI cells, validating the helicase domain as the primary drug target. Fragment-based screening led to the development of potent and highly selective WRN helicase covalent inhibitors. These compounds selectively suppressed MSI model growth In vitro and In vivo by mimicking WRN loss, inducing DNA double-strand breaks at expanded TA-repeats and DNA damage. Assessment of biomarkers in preclinical models linked TA-repeat expansions and mismatch repair (MMR) alterations to compound activity. Efficacy was confirmed in immunotherapy-resistant organoids and patient-derived xenograft (PDX) models. The discovery of potent, selective covalent WRN inhibitors provides proof of concept for synthetic-lethal targeting of WRN in MSI cancer and tools to dissect WRN biology.

Base editing screens map mutations affecting interferon-γ signaling in cancer.

Interferon-γ (IFN-γ) signaling mediates host responses to infection, inflammation and anti-tumor immunity. Mutations in the IFN-γ signaling pathway cause immunological disorders, hematological malignancies, and resistance to immune checkpoint blockade (ICB) in cancer; however, the function of most clinically observed variants remains unknown. Here, we systematically investigate the genetic determinants of IFN-γ response in colorectal cancer cells using CRISPR-Cas9 screens and base editing mutagenesis. Deep mutagenesis of JAK1 with cytidine and adenine base editors, combined with pathway-wide screens, reveal loss-of-function and gain-of-function mutations, including causal variants in hematological malignancies and mutations detected in patients refractory to ICB. We functionally validate variants of uncertain significance in primary tumor organoids, where engineering missense mutations in JAK1 enhanced or reduced sensitivity to autologous tumor-reactive T cells. We identify more than 300 predicted missense mutations altering IFN-γ pathway activity, generating a valuable resource for interpreting gene variant function.

Learning Insurance Benefit Rules from Policy Texts with Small Labeled Data.

To protect vital health program funds from being paid out on services that are wasteful and inconsistent with medical practices, government healthcare insurance programs need to validate the integrity of claims submitted by providers for reimbursement. However, due the complexity of healthcare billing policies and the lack of coded rules, maintaining "integrity" is a labor-intensive task, often narrow-scope and expensive. We propose an approach that combines deep learning and an ontology to support the extraction of actionable knowledge on benefit rules from regulatory healthcare policy text. We demonstrate its feasibility even in the presence of small ground truth labeled data provided by policy investigators. Leveraging deep learning and rich ontological information enables the system to learn from human corrections and capture better benefit rules from policy text, beyond just using a deterministic approach based on pre-defined textual and semantic pattterns.

Werner Helicase Is a Synthetic-Lethal Vulnerability in Mismatch Repair-Deficient Colorectal Cancer Refractory to Targeted Therapies, Chemotherapy, and Immunotherapy.

Targeted therapies, chemotherapy, and immunotherapy are used to treat patients with mismatch repair-deficient (dMMR)/microsatellite instability-high (MSI-H) colorectal cancer. The clinical effectiveness of targeted therapy and chemotherapy is limited by resistance and drug toxicities, and about half of patients receiving immunotherapy have disease that is refractory to immune checkpoint inhibitors. Loss of Werner syndrome ATP-dependent helicase (WRN) is a synthetic lethality in dMMR/MSI-H cells. To inform the development of WRN as a therapeutic target, we performed WRN knockout or knockdown in 60 heterogeneous dMMR colorectal cancer preclinical models, demonstrating that WRN dependency is an almost universal feature and a robust marker for patient selection. Furthermore, models of resistance to clinically relevant targeted therapy, chemotherapy, and immunotherapy retain WRN dependency. These data show the potential of therapeutically targeting WRN in patients with dMMR/MSI-H colorectal cancer and support WRN as a therapeutic option for patients with dMMR/MSI-H cancers refractory to current treatment strategies. SIGNIFICANCE: We found that a large, diverse set of dMMR/MSI-H colorectal cancer preclinical models, including models of treatment-refractory disease, are WRN-dependent. Our results support WRN as a promising synthetic-lethal target in dMMR/MSI-H colorectal cancer tumors as a monotherapy or in combination with targeted agents, chemotherapy, or immunotherapy.This article is highlighted in the In This Issue feature, p. 1861.

Minimal genome-wide human CRISPR-Cas9 library.

CRISPR guide RNA libraries have been iteratively improved to provide increasingly efficient reagents, although their large size is a barrier for many applications. We design an optimised minimal genome-wide human CRISPR-Cas9 library (MinLibCas9) by mining existing large-scale gene loss-of-function datasets, resulting in a greater than 42% reduction in size compared to other CRISPR-Cas9 libraries while preserving assay sensitivity and specificity. MinLibCas9 provides backward compatibility with existing datasets, increases the dynamic range of CRISPR-Cas9 screens and extends their application to complex models and assays.

Drug mechanism-of-action discovery through the integration of pharmacological and CRISPR screens.

Low success rates during drug development are due, in part, to the difficulty of defining drug mechanism-of-action and molecular markers of therapeutic activity. Here, we integrated 199,219 drug sensitivity measurements for 397 unique anti-cancer drugs with genome-wide CRISPR loss-of-function screens in 484 cell lines to systematically investigate cellular drug mechanism-of-action. We observed an enrichment for positive associations between the profile of drug sensitivity and knockout of a drug's nominal target, and by leveraging protein-protein networks, we identified pathways underpinning drug sensitivity. This revealed an unappreciated positive association between mitochondrial E3 ubiquitin-protein ligase MARCH5 dependency and sensitivity to MCL1 inhibitors in breast cancer cell lines. We also estimated drug on-target and off-target activity, informing on specificity, potency and toxicity. Linking drug and gene dependency together with genomic data sets uncovered contexts in which molecular networks when perturbed mediate cancer cell loss-of-fitness and thereby provide independent and orthogonal evidence of biomarkers for drug development. This study illustrates how integrating cell line drug sensitivity with CRISPR loss-of-function screens can elucidate mechanism-of-action to advance drug development.

Improved Outcome Prediction for Appendiceal Pseudomyxoma Peritonei by Integration of Cancer Cell and Stromal Transcriptional Profiles.

In recent years, cytoreductive surgery (CRS) and hyperthermic intraperitoneal chemotherapy (HIPEC) have substantially improved the clinical outcome of pseudomyxoma peritonei (PMP) originating from mucinous appendiceal cancer. However, current histopathological grading of appendiceal PMP frequently fails in predicting disease outcome. We recently observed that the integration of cancer cell transcriptional traits with those of cancer-associated fibroblasts (CAFs) improves prognostic prediction for tumors of the large intestine. We therefore generated global expression profiles on a consecutive series of 24 PMP patients treated with CRS plus HIPEC. Multiple lesions were profiled for nine patients. We then used expression data to stratify the samples by a previously published "high-risk appendiceal cancer" (HRAC) signature and by a CAF signature that we previously developed for colorectal cancer, or by a combination of both. The prognostic value of the HRAC signature was confirmed in our cohort and further improved by integration of the CAF signature. Classification of cases profiled for multiple lesions revealed the existence of outlier samples and highlighted the need of profiling multiple PMP lesions to select representative samples for optimal performances. The integrated predictor was subsequently validated in an independent PMP cohort. These results provide new insights into PMP biology, revealing a previously unrecognized prognostic role of the stromal component and supporting integration of standard pathological grade with the HRAC and CAF transcriptional signatures to better predict disease outcome.

Ontology-Guided Policy Information Extraction for Healthcare Fraud Detection.

Financial losses in Medicaid, from Fraud, Waste and Abuse (FWA), in the United States are estimated to be in the tens of billions of dollars each year. This results in escalating costs as well as limiting the funding available to worthy recipients of healthcare. The Centers for Medicare & Medicaid Services mandate thorough auditing, in which policy investigators manually research and interpret the policy to validate the integrity of claims submitted by providers for reimbursement, a very time-consuming process. We propose a system that aims to interpret unstructured policy text to semi-automatically audit provider claims. Guided by a domain ontology, our system extracts entities and relations to build benefit rules that can be executed on top of claims to identify improper payments, and often in turn payment policy or claims adjudication system vulnerabilities. We validate the automatic knowledge extraction from policies based on ground truth created by domain experts. Lastly, we discuss how the system can co-reason with human investigators in order to increase thoroughness and consistency in the review of claims and policy, to identify providers that systematically violate policies and to help in prioritising investigations.

Patient-Derived Xenografts and Matched Cell Lines Identify Pharmacogenomic Vulnerabilities in Colorectal Cancer.

PURPOSE: Patient-derived xenograft (PDX) models accurately recapitulate the tumor of origin in terms of histopathology, genomic landscape, and therapeutic response, but some limitations due to costs associated with their maintenance and restricted amenability for large-scale screenings still exist. To overcome these issues, we established a platform of 2D cell lines (xeno-cell lines, XL), derived from PDXs of colorectal cancer with matched patient germline gDNA available. EXPERIMENTAL DESIGN: Whole-exome and transcriptome sequencing analyses were performed. Biomarkers of response and resistance to anti-HER therapy were annotated. Dependency on the WRN helicase gene was assessed in MSS, MSI-H, and MSI-like XLs using a reverse genetics functional approach. RESULTS: XLs recapitulated the entire spectrum of colorectal cancer transcriptional subtypes. Exome and RNA-seq analyses delineated several molecular biomarkers of response and resistance to EGFR and HER2 blockade. Genotype-driven responses observed in vitro in XLs were confirmed in vivo in the matched PDXs. MSI-H models were dependent upon WRN gene expression, while loss of WRN did not affect MSS XLs growth. Interestingly, one MSS XL with transcriptional MSI-like traits was sensitive to WRN depletion. CONCLUSIONS: The XL platform represents a preclinical tool for functional gene validation and proof-of-concept studies to identify novel druggable vulnerabilities in colorectal cancer.

Functional linkage of gene fusions to cancer cell fitness assessed by pharmacological and CRISPR-Cas9 screening.

Many gene fusions are reported in tumours and for most their role remains unknown. As fusions are used for diagnostic and prognostic purposes, and are targets for treatment, it is crucial to assess their function in cancer. To systematically investigate the role of fusions in tumour cell fitness, we utilized RNA-sequencing data from 1011 human cancer cell lines to functionally link 8354 fusion events with genomic data, sensitivity to >350 anti-cancer drugs and CRISPR-Cas9 loss-of-fitness effects. Established clinically-relevant fusions were identified. Overall, detection of functional fusions was rare, including those involving cancer driver genes, suggesting that many fusions are dispensable for tumour fitness. Therapeutically actionable fusions involving RAF1, BRD4 and ROS1 were verified in new histologies. In addition, recurrent YAP1-MAML2 fusions were identified as activators of Hippo-pathway signaling in multiple cancer types. Our approach discriminates functional fusions, identifying new drivers of carcinogenesis and fusions that could have clinical implications.

Prioritization of cancer therapeutic targets using CRISPR-Cas9 screens.

Functional genomics approaches can overcome limitations-such as the lack of identification of robust targets and poor clinical efficacy-that hamper cancer drug development. Here we performed genome-scale CRISPR-Cas9 screens in 324 human cancer cell lines from 30 cancer types and developed a data-driven framework to prioritize candidates for cancer therapeutics. We integrated cell fitness effects with genomic biomarkers and target tractability for drug development to systematically prioritize new targets in defined tissues and genotypes. We verified one of our most promising dependencies, the Werner syndrome ATP-dependent helicase, as a synthetic lethal target in tumours from multiple cancer types with microsatellite instability. Our analysis provides a resource of cancer dependencies, generates a framework to prioritize cancer drug targets and suggests specific new targets. The principles described in this study can inform the initial stages of drug development by contributing to a new, diverse and more effective portfolio of cancer drug targets.

A Road Map for Precision Cancer Medicine Using Personalized Models.

A study by Pauli and colleagues in this issue of Cancer Discovery describes the creation of a precision cancer platform for patients with advanced disease, integrating DNA sequencing of patient tumors with the generation of patient-derived organoids and xenografts. They propose the use of this platform for drug testing to nominate therapeutic options for individual patients and for therapeutic biomarker discovery. Cancer Discov; 7(5); 456-8. ©2017 AACRSee related article by Pauli et al., p. 462.

Loss of AXIN1 drives acquired resistance to WNT pathway blockade in colorectal cancer cells carrying RSPO3 fusions.

In colorectal cancer (CRC), WNT pathway activation by genetic rearrangements of RSPO3 is emerging as a promising target. However, its low prevalence severely limits availability of preclinical models for in-depth characterization. Using a pipeline designed to suppress stroma-derived signal, we find that RSPO3 "outlier" expression in CRC samples highlights translocation and fusion transcript expression. Outlier search in 151 CRC cell lines identified VACO6 and SNU1411 cells as carriers of, respectively, a canonical PTPRK(e1)-RSPO3(e2) fusion and a novel PTPRK(e13)-RSPO3(e2) fusion. Both lines displayed marked in vitro and in vivo sensitivity to WNT blockade by the porcupine inhibitor LGK974, associated with transcriptional and morphological evidence of WNT pathway suppression. Long-term treatment of VACO6 cells with LGK974 led to the emergence of a resistant population carrying two frameshift deletions of the WNT pathway inhibitor AXIN1, with consequent protein loss. Suppression of AXIN1 in parental VACO6 cells by RNA interference conferred marked resistance to LGK974. These results provide the first mechanism of secondary resistance to WNT pathway inhibition.

The RNA-binding protein ESRP1 promotes human colorectal cancer progression.

Epithelial splicing regulatory protein 1 (ESRP1) is an epithelial cell-specific RNA binding protein that controls several key cellular processes, like alternative splicing and translation. Previous studies have demonstrated a tumor suppressor role for this protein. Recently, however, a pro-metastatic function of ESRP1 has been reported. We thus aimed at clarifying the role of ESRP1 in Colorectal Cancer (CRC) by performing loss- and gain-of-function studies, and evaluating tumorigenesis and malignancy with in vitro and in vivo approaches. We found that ESRP1 plays a role in anchorage-independent growth of CRC cells. ESRP1-overexpressing cells grown in suspension showed enhanced fibroblast growth factor receptor (FGFR1/2) signalling, Akt activation, and Snail upregulation. Moreover, ESRP1 promoted the ability of CRC cells to generate macrometastases in mice livers. High ESRP1 expression may thus stimulate growth of cancer epithelial cells and promote colorectal cancer progression. Our findings provide mechanistic insights into a previously unreported, pro-oncogenic role for ESRP1 in CRC, and suggest that fine-tuning the level of this RNA-binding protein could be relevant in modulating tumor growth in a subset of CRC patients.

Efficacy of NEDD8 Pathway Inhibition in Preclinical Models of Poorly Differentiated, Clinically Aggressive Colorectal Cancer.

Background: The NEDD8 conjugation pathway modulates the ubiquitination and activity of a wide range of intracellular proteins, and its blockade by pevonedistat is emerging as a promising therapeutic approach in various cancer settings. However, systematic characterization of pevonedistat efficacy in specific tumor types and definition of response predictors are still missing. Methods: We investigated in vitro sensitivity to pevonedistat in 122 colorectal cancer (CRC) cell lines by an ATP-based proliferation assay and evaluated apoptosis and DNA content by flow cytometry. Associations between pevonedistat sensitivity and CRC molecular features were assessed by Student's t test. A 184-gene transcriptional predictor was generated in cell lines and applied to 87 metastatic CRC samples for which patient-derived xenografts (PDXs) were available. In vivo reponse to pevonedistat was assessed in PDX models (≥5 mice per group). All statistical tests were two-sided. Results: Sixteen (13.1%) cell lines displayed a marked response to pevonedistat, featuring DNA re-replication, proliferative block, and increased apoptosis. Pevonedistat sensitivity did not statistically significantly correlate with microsatellite instability or mutations in KRAS or BRAF and was functionally associated with low EGFR pathway activity. While ineffective on predicted resistant PDXs, in vivo administration of pevonedistat statistically significantly impaired growth of five out of six predicted sensitive models (P < .01). In samples from CRC patients, transcriptional prediction of pevonedistat sensitivity was associated with poor prognosis after surgery (hazard ratio [HR] = 2.49, 95% confidence interval [CI] = 1.34 to 4.62, P = .003) and early progression under cetuximab treatment (HR = 3.59, 95% CI = 1.60 to 8.04, P < .001). Histological and immunohistochemical analyses revealed that the pevonedistat sensitivity signature captures transcriptional traits of poor differentiation and high-grade mucinous adenocarcinoma. Conclusions: These results highlight NEDD8-pathway inhibition by pevonedistat as a potentially effective treatment for poorly differentiated, clinically aggressive CRC.

A diphtheria toxin resistance marker for in vitro and in vivo selection of stably transduced human cells.

We developed a selectable marker rendering human cells resistant to Diphtheria Toxin (DT). The marker (DT(R)) consists of a primary microRNA sequence engineered to downregulate the ubiquitous DPH2 gene, a key enzyme for the biosynthesis of the DT target diphthamide. DT(R) expression in human cells invariably rendered them resistant to DT in vitro, without altering basal cell growth. DT(R)-based selection efficiency and stability were comparable to those of established drug-resistance markers. As mice are insensitive to DT, DT(R)-based selection can be also applied in vivo. Direct injection of a GFP-DT(R) lentiviral vector into human cancer cell-line xenografts and patient-derived tumorgrafts implanted in mice, followed by systemic DT administration, yielded tumors entirely composed of permanently transduced cells and detectable by imaging systems. This approach enabled high-efficiency in vivo selection of xenografted human tumor tissues expressing ectopic transgenes, a hitherto unmet need for functional and morphological studies in laboratory animals.

The molecular landscape of colorectal cancer cell lines unveils clinically actionable kinase targets.

The development of molecularly targeted anticancer agents relies on large panels of tumour-specific preclinical models closely recapitulating the molecular heterogeneity observed in patients. Here we describe the mutational and gene expression analyses of 151 colorectal cancer (CRC) cell lines. We find that the whole spectrum of CRC molecular and transcriptional subtypes, previously defined in patients, is represented in this cell line compendium. Transcriptional outlier analysis identifies RAS/BRAF wild-type cells, resistant to EGFR blockade, functionally and pharmacologically addicted to kinase genes including ALK, FGFR2, NTRK1/2 and RET. The same genes are present as expression outliers in CRC patient samples. Genomic rearrangements (translocations) involving the ALK and NTRK1 genes are associated with the overexpression of the corresponding proteins in CRC specimens. The approach described here can be used to pinpoint CRCs with exquisite dependencies to individual kinases for which clinically approved drugs are already available.

MicroRNA-mRNA interactions underlying colorectal cancer molecular subtypes.

Colorectal cancer (CRC) transcriptional subtypes have been recently identified by gene expression profiling. Here we describe an analytical pipeline, microRNA master regulator analysis (MMRA), developed to search for microRNAs potentially driving CRC subtypes. Starting from a microRNA-mRNA tumour expression data set, MMRA identifies candidate regulator microRNAs by assessing their subtype-specific expression, target enrichment in subtype mRNA signatures and network analysis-based contribution to subtype gene expression. When applied to a CRC data set of 450 samples, assigned to subtypes by 3 different transcriptional classifiers, MMRA identifies 24 candidate microRNAs, in most cases downregulated in the stem/serrated/mesenchymal (SSM) poor prognosis subtype. Functional validation in CRC cell lines confirms downregulation of the SSM subtype by miR-194, miR-200b, miR-203 and miR-429, which share target genes and pathways mediating this effect. These results show that, by combining statistical tests, target prediction and network analysis, MMRA effectively identifies microRNAs functionally associated to cancer subtypes.

Truncated RAF kinases drive resistance to MET inhibition in MET-addicted cancer cells.

Constitutively active receptor tyrosine kinases (RTKs) are known oncogenic drivers and provide valuable therapeutic targets in many cancer types. However, clinical efficacy of RTK inhibitors is limited by intrinsic and acquired resistance. To identify genes conferring resistance to inhibition of the MET RTK, we conducted a forward genetics screen in the GTL-16 gastric cancer cell line, carrying MET amplification and exquisitely sensitive to MET inhibition. Cells were transduced with three different retroviral cDNA expression libraries and selected for growth in the presence of the MET inhibitor PHA-665752. Selected cells displayed robust and reproducible enrichment of library-derived cDNAs encoding truncated forms of RAF1 and BRAF proteins, whose silencing reversed the resistant phenotype. Transduction of naïve GTL-16 cells with truncated, but not full length, RAF1 and BRAF conferred in vitro and in vivo resistance to MET inhibitors, which could be reversed by MEK inhibition. Induction of resistance by truncated RAFs was confirmed in other MET-addicted cell lines, and further extended to EGFR-addicted cells. These data show that truncated RAF1 and BRAF proteins, recently described as products of genomic rearrangements in gastric cancer and other malignancies, have the ability to render neoplastic cells resistant to RTK-targeted therapy.