Machine learning in rare disease.

High-throughput profiling methods (such as genomics or imaging) have accelerated basic research and made deep molecular characterization of patient samples routine. These approaches provide a rich portrait of genes, molecular pathways and cell types involved in disease phenotypes. Machine learning (ML) can be a useful tool for extracting disease-relevant patterns from high-dimensional datasets. However, depending upon the complexity of the biological question, machine learning often requires many samples to identify recurrent and biologically meaningful patterns. Rare diseases are inherently limited in clinical cases, leading to few samples to study. In this Perspective, we outline the challenges and emerging solutions for using ML for small sample sets, specifically in rare diseases. Advances in ML methods for rare diseases are likely to be informative for applications beyond rare diseases for which few samples exist with high-dimensional data. We propose that the method community prioritize the development of ML techniques for rare disease research.

COVID-19 in people with neurofibromatosis 1, neurofibromatosis 2, or schwannomatosis.

PURPOSE: People with pre-existing conditions may be more susceptible to severe COVID-19 when infected by SARS-CoV-2. The relative risk and severity of SARS-CoV-2 infection in people with rare diseases such as neurofibromatosis type 1 (NF1), neurofibromatosis type 2 (NF2), or schwannomatosis (SWN) is unknown. METHODS: We investigated the proportions of people with NF1, NF2, or SWN in the National COVID Cohort Collaborative (N3C) electronic health record data set who had a positive test result for SARS-CoV-2 or COVID-19. RESULTS: The cohort sizes in N3C were 2501 (NF1), 665 (NF2), and 762 (SWN). We compared these with N3C cohorts of patients with other rare diseases (98-9844 individuals) and the general non-NF population of 5.6 million. The site- and age-adjusted proportion of people with NF1, NF2, or SWN who had a positive test result for SARS-CoV-2 or COVID-19 (collectively termed positive cases) was not significantly higher than in individuals without NF or other selected rare diseases. There were no severe outcomes reported in the NF2 or SWN cohorts. The proportion of patients experiencing severe outcomes was no greater for people with NF1 than in cohorts with other rare diseases or the general population. CONCLUSION: Having NF1, NF2, or SWN does not appear to increase the risk of being SARS-CoV-2 positive or of being a patient with COVID-19 or of developing severe complications from SARS-CoV-2.

A Crowdsourcing Approach to Develop Machine Learning Models to Quantify Radiographic Joint Damage in Rheumatoid Arthritis.

Importance: An automated, accurate method is needed for unbiased assessment quantifying accrual of joint space narrowing and erosions on radiographic images of the hands and wrists, and feet for clinical trials, monitoring of joint damage over time, assisting rheumatologists with treatment decisions. Such a method has the potential to be directly integrated into electronic health records. Objectives: To design and implement an international crowdsourcing competition to catalyze the development of machine learning methods to quantify radiographic damage in rheumatoid arthritis (RA). Design, Setting, and Participants: This diagnostic/prognostic study describes the Rheumatoid Arthritis 2-Dialogue for Reverse Engineering Assessment and Methods (RA2-DREAM Challenge), which used existing radiographic images and expert-curated Sharp-van der Heijde (SvH) scores from 2 clinical studies (674 radiographic sets from 562 patients) for training (367 sets), leaderboard (119 sets), and final evaluation (188 sets). Challenge participants were tasked with developing methods to automatically quantify overall damage (subchallenge 1), joint space narrowing (subchallenge 2), and erosions (subchallenge 3). The challenge was finished on June 30, 2020. Main Outcomes and Measures: Scores derived from submitted algorithms were compared with the expert-curated SvH scores, and a baseline model was created for benchmark comparison. Performances were ranked using weighted root mean square error (RMSE). The performance and reproductivity of each algorithm was assessed using Bayes factor from bootstrapped data, and further evaluated with a postchallenge independent validation data set. Results: The RA2-DREAM Challenge received a total of 173 submissions from 26 participants or teams in 7 countries for the leaderboard round, and 13 submissions were included in the final evaluation. The weighted RMSEs metric showed that the winning algorithms produced scores that were very close to the expert-curated SvH scores. Top teams included Team Shirin for subchallenge 1 (weighted RMSE, 0.44), HYL-YFG (Hongyang Li and Yuanfang Guan) subchallenge 2 (weighted RMSE, 0.38), and Gold Therapy for subchallenge 3 (weighted RMSE, 0.43). Bootstrapping/Bayes factor approach and the postchallenge independent validation confirmed the reproducibility and the estimation concordance indices between final evaluation and postchallenge independent validation data set were 0.71 for subchallenge 1, 0.78 for subchallenge 2, and 0.82 for subchallenge 3. Conclusions and Relevance: The RA2-DREAM Challenge resulted in the development of algorithms that provide feasible, quick, and accurate methods to quantify joint damage in RA. Ultimately, these methods could help research studies on RA joint damage and may be integrated into electronic health records to help clinicians serve patients better by providing timely, reliable, and quantitative information for making treatment decisions to prevent further damage.

Centralizing neurofibromatosis experimental tool knowledge with the NF Research Tools Database.

Experimental tools and resources, such as animal models, cell lines, antibodies, genetic reagents and biobanks, are key ingredients in biomedical research. Investigators face multiple challenges when trying to understand the availability, applicability and accessibility of these tools. A major challenge is keeping up with current information about the numerous tools available for a particular research problem. A variety of disease-agnostic projects such as the Mouse Genome Informatics database and the Resource Identification Initiative curate a number of types of research tools. Here, we describe our efforts to build upon these resources to develop a disease-specific research tool resource for the neurofibromatosis (NF) research community. This resource, the NF Research Tools Database, is an open-access database that enables the exploration and discovery of information about NF type 1-relevant animal models, cell lines, antibodies, genetic reagents and biobanks. Users can search and explore tools, obtain detailed information about each tool as well as read and contribute their observations about the performance, reliability and characteristics of tools in the database. NF researchers will be able to use the NF Research Tools Database to promote, discover, share, reuse and characterize research tools, with the goal of advancing NF research. Database URL: https://tools.nf.synapse.org/.

A community challenge for a pancancer drug mechanism of action inference from perturbational profile data.

The Columbia Cancer Target Discovery and Development (CTD2) Center is developing PANACEA, a resource comprising dose-responses and RNA sequencing (RNA-seq) profiles of 25 cell lines perturbed with ∼400 clinical oncology drugs, to study a tumor-specific drug mechanism of action. Here, this resource serves as the basis for a DREAM Challenge assessing the accuracy and sensitivity of computational algorithms for de novo drug polypharmacology predictions. Dose-response and perturbational profiles for 32 kinase inhibitors are provided to 21 teams who are blind to the identity of the compounds. The teams are asked to predict high-affinity binding targets of each compound among ∼1,300 targets cataloged in DrugBank. The best performing methods leverage gene expression profile similarity analysis as well as deep-learning methodologies trained on individual datasets. This study lays the foundation for future integrative analyses of pharmacogenomic data, reconciliation of polypharmacology effects in different tumor contexts, and insights into network-based assessments of drug mechanisms of action.

Crowdsourced identification of multi-target kinase inhibitors for RET- and TAU- based disease: The Multi-Targeting Drug DREAM Challenge.

A continuing challenge in modern medicine is the identification of safer and more efficacious drugs. Precision therapeutics, which have one molecular target, have been long promised to be safer and more effective than traditional therapies. This approach has proven to be challenging for multiple reasons including lack of efficacy, rapidly acquired drug resistance, and narrow patient eligibility criteria. An alternative approach is the development of drugs that address the overall disease network by targeting multiple biological targets ('polypharmacology'). Rational development of these molecules will require improved methods for predicting single chemical structures that target multiple drug targets. To address this need, we developed the Multi-Targeting Drug DREAM Challenge, in which we challenged participants to predict single chemical entities that target pro-targets but avoid anti-targets for two unrelated diseases: RET-based tumors and a common form of inherited Tauopathy. Here, we report the results of this DREAM Challenge and the development of two neural network-based machine learning approaches that were applied to the challenge of rational polypharmacology. Together, these platforms provide a potentially useful first step towards developing lead therapeutic compounds that address disease complexity through rational polypharmacology.

Crowdsourced mapping of unexplored target space of kinase inhibitors.

Despite decades of intensive search for compounds that modulate the activity of particular protein targets, a large proportion of the human kinome remains as yet undrugged. Effective approaches are therefore required to map the massive space of unexplored compound-kinase interactions for novel and potent activities. Here, we carry out a crowdsourced benchmarking of predictive algorithms for kinase inhibitor potencies across multiple kinase families tested on unpublished bioactivity data. We find the top-performing predictions are based on various models, including kernel learning, gradient boosting and deep learning, and their ensemble leads to a predictive accuracy exceeding that of single-dose kinase activity assays. We design experiments based on the model predictions and identify unexpected activities even for under-studied kinases, thereby accelerating experimental mapping efforts. The open-source prediction algorithms together with the bioactivities between 95 compounds and 295 kinases provide a resource for benchmarking prediction algorithms and for extending the druggable kinome.

A clinically and genomically annotated nerve sheath tumor biospecimen repository.

Nerve sheath tumors occur as a heterogeneous group of neoplasms in patients with neurofibromatosis type 1 (NF1). The malignant form represents the most common cause of death in people with NF1, and even when benign, these tumors can result in significant disfigurement, neurologic dysfunction, and a range of profound symptoms. Lack of human tissue across the peripheral nerve tumors common in NF1 has been a major limitation in the development of new therapies. To address this unmet need, we have created an annotated collection of patient tumor samples, patient-derived cell lines, and patient-derived xenografts, and carried out high-throughput genomic and transcriptomic characterization to serve as a resource for further biologic and preclinical therapeutic studies. In this work, we release genomic and transcriptomic datasets comprised of 55 tumor samples derived from 23 individuals, complete with clinical annotation. All data are publicly available through the NF Data Portal and at http://synapse.org/jhubiobank.

Integrative Analysis Identifies Candidate Tumor Microenvironment and Intracellular Signaling Pathways that Define Tumor Heterogeneity in NF1.

Neurofibromatosis type 1 (NF1) is a monogenic syndrome that gives rise to numerous symptoms including cognitive impairment, skeletal abnormalities, and growth of benign nerve sheath tumors. Nearly all NF1 patients develop cutaneous neurofibromas (cNFs), which occur on the skin surface, whereas 40-60% of patients develop plexiform neurofibromas (pNFs), which are deeply embedded in the peripheral nerves. Patients with pNFs have a ~10% lifetime chance of these tumors becoming malignant peripheral nerve sheath tumors (MPNSTs). These tumors have a severe prognosis and few treatment options other than surgery. Given the lack of therapeutic options available to patients with these tumors, identification of druggable pathways or other key molecular features could aid ongoing therapeutic discovery studies. In this work, we used statistical and machine learning methods to analyze 77 NF1 tumors with genomic data to characterize key signaling pathways that distinguish these tumors and identify candidates for drug development. We identified subsets of latent gene expression variables that may be important in the identification and etiology of cNFs, pNFs, other neurofibromas, and MPNSTs. Furthermore, we characterized the association between these latent variables and genetic variants, immune deconvolution predictions, and protein activity predictions.

Spatiotemporal Loss of NF1 in Schwann Cell Lineage Leads to Different Types of Cutaneous Neurofibroma Susceptible to Modification by the Hippo Pathway.

Neurofibromatosis type 1 (NF1) is a cancer predisposition disorder that results from inactivation of the tumor suppressor neurofibromin, a negative regulator of RAS signaling. Patients with NF1 present with a wide range of clinical manifestations, and the tumor with highest prevalence is cutaneous neurofibroma (cNF). Most patients harboring cNF suffer greatly from the burden of those tumors, which have no effective medical treatment. Ironically, none of the numerous NF1 mouse models developed so far recapitulate cNF. Here, we discovered that HOXB7 serves as a lineage marker to trace the developmental origin of cNF neoplastic cells. Ablating Nf1 in the HOXB7 lineage faithfully recapitulates both human cutaneous and plexiform neurofibroma. In addition, we discovered that modulation of the Hippo pathway acts as a "modifier" for neurofibroma tumorigenesis. This mouse model opens the doors for deciphering the evolution of cNF to identify effective therapies, where none exist today. SIGNIFICANCE: This study provides insights into the developmental origin of cNF, the most common tumor in NF1, and generates the first mouse model that faithfully recapitulates both human cutaneous and plexiform neurofibroma. The study also demonstrates that the Hippo pathway can modify neurofibromagenesis, suggesting that dampening the Hippo pathway could be an attractive therapeutic target.This article is highlighted in the In This Issue feature, p. 1.

Probing the chemical-biological relationship space with the Drug Target Explorer.

Modern phenotypic high-throughput screens (HTS) present several challenges including identifying the target(s) that mediate the effect seen in the screen, characterizing 'hits' with a polypharmacologic target profile, and contextualizing screen data within the large space of drugs and screening models. To address these challenges, we developed the Drug-Target Explorer. This tool allows users to query molecules within a database of experimentally-derived and curated compound-target interactions to identify structurally similar molecules and their targets. It enables network-based visualizations of the compound-target interaction space, and incorporates comparisons to publicly-available in vitro HTS datasets. Furthermore, users can identify molecules using a query target or set of targets. The Drug Target Explorer is a multifunctional platform for exploring chemical space as it relates to biological targets, and may be useful at several steps along the drug development pipeline including target discovery, structure-activity relationship, and lead compound identification studies.

Traditional and systems biology based drug discovery for the rare tumor syndrome neurofibromatosis type 2.

Neurofibromatosis 2 (NF2) is a rare tumor suppressor syndrome that manifests with multiple schwannomas and meningiomas. There are no effective drug therapies for these benign tumors and conventional therapies have limited efficacy. Various model systems have been created and several drug targets have been implicated in NF2-driven tumorigenesis based on known effects of the absence of merlin, the product of the NF2 gene. We tested priority compounds based on known biology with traditional dose-concentration studies in meningioma and schwann cell systems. Concurrently, we studied functional kinome and gene expression in these cells pre- and post-treatment to determine merlin deficient molecular phenotypes. Cell viability results showed that three agents (GSK2126458, Panobinostat, CUDC-907) had the greatest activity across schwannoma and meningioma cell systems, but merlin status did not significantly influence response. In vivo, drug effect was tumor specific with meningioma, but not schwannoma, showing response to GSK2126458 and Panobinostat. In culture, changes in both the transcriptome and kinome in response to treatment clustered predominantly based on tumor type. However, there were differences in both gene expression and functional kinome at baseline between meningioma and schwannoma cell systems that may form the basis for future selective therapies. This work has created an openly accessible resource (www.synapse.org/SynodosNF2) of fully characterized isogenic schwannoma and meningioma cell systems as well as a rich data source of kinome and transcriptome data from these assay systems before and after treatment that enables single and combination drug discovery based on molecular phenotype.

Exploiting mitochondrial and metabolic homeostasis as a vulnerability in NF1 deficient cells.

Neurofibromatosis type 1 is a disease caused by mutation of neurofibromin 1 (NF1), loss of which results in hyperactive Ras signaling and a concomitant increase in cell proliferation and survival. Patients with neurofibromatosis type 1 frequently develop tumors such as plexiform neurofibromas and malignant peripheral nerve sheath tumors. Mutation of NF1 or loss of the NF1 protein is also observed in glioblastoma, lung adenocarcinoma, and ovarian cancer among other sporadic cancers. A therapy that selectively targets NF1 deficient tumors would substantially advance our ability to treat these malignancies. To address the need for these therapeutics, we developed and conducted a synthetic lethality screen to discover molecules that target yeast lacking the homolog of NF1, IRA2. One of the lead candidates that was observed to be synthetic lethal with ira2Δ yeast is Y100. Here, we describe the mechanisms by which Y100 targets ira2Δ yeast and NF1-deficient tumor cells. Y100 treatment disrupted proteostasis, metabolic homeostasis, and induced the formation of mitochondrial superoxide in NF1-deficient cancer cells. Previous studies also indicate that NF1/Ras-dysregulated tumors may be sensitive to modulators of oxidative and ER stress. We hypothesize that the use of Y100 and molecules with related mechanisms of action represent a feasible therapeutic strategy for targeting NF1 deficient cells.

Cutaneous neurofibromas in the genomics era: current understanding and open questions.

Cutaneous neurofibromas (cNF) are a nearly ubiquitous symptom of neurofibromatosis type 1 (NF1), a disorder with a broad phenotypic spectrum caused by germline mutation of the neurofibromatosis type 1 tumour suppressor gene (NF1). Symptoms of NF1 can include learning disabilities, bone abnormalities and predisposition to tumours such as cNFs, plexiform neurofibromas, malignant peripheral nerve sheath tumours and optic nerve tumours. There are no therapies currently approved for cNFs aside from elective surgery, and the molecular aetiology of cNF remains relatively uncharacterised. Furthermore, whereas the biallelic inactivation of NF1 in neoplastic Schwann cells is critical for cNF formation, it is still unclear which additional genetic, transcriptional, epigenetic, microenvironmental or endocrine changes are important. Significant inroads have been made into cNF understanding, including NF1 genotype-phenotype correlations in NF1 microdeletion patients, the identification of recurring somatic mutations, studies of cNF-invading mast cells and macrophages, and clinical trials of putative therapeutic targets such as mTOR, MEK and c-KIT. Despite these advances, several gaps remain in our knowledge of the associated pathogenesis, which is further hampered by a lack of translationally relevant animal models. Some of these questions may be addressed in part by the adoption of genomic analysis techniques. Understanding the aetiology of cNF at the genomic level may assist in the development of new therapies for cNF, and may also contribute to a greater understanding of NF1/RAS signalling in cancers beyond those associated with NF1. Here, we summarise the present understanding of cNF biology, including the pathogenesis, mutational landscape, contribution of the tumour microenvironment and endocrine signalling, and the historical and current state of clinical trials for cNF. We also highlight open access data resources and potential avenues for future research that leverage recently developed genomics-based methods in cancer research.

A machine learning classifier trained on cancer transcriptomes detects NF1 inactivation signal in glioblastoma.

BACKGROUND: We have identified molecules that exhibit synthetic lethality in cells with loss of the neurofibromin 1 (NF1) tumor suppressor gene. However, recognizing tumors that have inactivation of the NF1 tumor suppressor function is challenging because the loss may occur via mechanisms that do not involve mutation of the genomic locus. Degradation of the NF1 protein, independent of NF1 mutation status, phenocopies inactivating mutations to drive tumors in human glioma cell lines. NF1 inactivation may alter the transcriptional landscape of a tumor and allow a machine learning classifier to detect which tumors will benefit from synthetic lethal molecules. RESULTS: We developed a strategy to predict tumors with low NF1 activity and hence tumors that may respond to treatments that target cells lacking NF1. Using RNAseq data from The Cancer Genome Atlas (TCGA), we trained an ensemble of 500 logistic regression classifiers that integrates mutation status with whole transcriptomes to predict NF1 inactivation in glioblastoma (GBM). On TCGA data, the classifier detected NF1 mutated tumors (test set area under the receiver operating characteristic curve (AUROC) mean = 0.77, 95% quantile = 0.53 - 0.95) over 50 random initializations. On RNA-Seq data transformed into the space of gene expression microarrays, this method produced a classifier with similar performance (test set AUROC mean = 0.77, 95% quantile = 0.53 - 0.96). We applied our ensemble classifier trained on the transformed TCGA data to a microarray validation set of 12 samples with matched RNA and NF1 protein-level measurements. The classifier's NF1 score was associated with NF1 protein concentration in these samples. CONCLUSIONS: We demonstrate that TCGA can be used to train accurate predictors of NF1 inactivation in GBM. The ensemble classifier performed well for samples with very high or very low NF1 protein concentrations but had mixed performance in samples with intermediate NF1 concentrations. Nevertheless, high-performing and validated predictors have the potential to be paired with targeted therapies and personalized medicine.

Genomic characterization of patient-derived xenograft models established from fine needle aspirate biopsies of a primary pancreatic ductal adenocarcinoma and from patient-matched metastatic sites.

N-of-1 trials target actionable mutations, yet such approaches do not test genomically-informed therapies in patient tumor models prior to patient treatment. To address this, we developed patient-derived xenograft (PDX) models from fine needle aspiration (FNA) biopsies (FNA-PDX) obtained from primary pancreatic ductal adenocarcinoma (PDAC) at the time of diagnosis. Here, we characterize PDX models established from one primary and two metastatic sites of one patient. We identified an activating KRAS G12R mutation among other mutations in these models. In explant cells derived from these PDX tumor models with a KRAS G12R mutation, treatment with inhibitors of CDKs (including CDK9) reduced phosphorylation of a marker of CDK9 activity (phospho-RNAPII CTD Ser2/5) and reduced viability/growth of explant cells derived from PDAC PDX models. Similarly, a CDK inhibitor reduced phospho-RNAPII CTD Ser2/5, increased apoptosis, and inhibited tumor growth in FNA-PDX and patient-matched metastatic-PDX models. In summary, PDX models can be constructed from FNA biopsies of PDAC which in turn can enable genomic characterization and identification of potential therapies.

Gold nanoparticle trafficking of typically excluded compounds across the cell membrane in JB6 Cl 41-5a cells causes assay interference.

Nanoparticles (NP) often interfere with the mechanism and interpretation of high throughput in vitro toxicity assays. This interference may occur at any time during the assay and spans most NP systems. This study reports on a specific type of gold NP assay interference, where unmodified gold NPs were able to traffic certain assay molecules that contained primary amines across the cell membrane resulting in false positive results for toxicity assays. The enhanced assay molecule permeability was eliminated when the gold NP surface was both sterically and chemically blocked by polyethylene glycol (PEG). The results support the growing consensus that appropriate controls and assay validation should occur prior to interpretation of results of assays using NP.