Citations

There is a growing interest in the application of machine learning (ML) in Alzheimer’s disease (AD) research. However, neuropsychiatric symptoms (NPS), frequent in subjects with AD, mild cognitive impairment (MCI), and other related dementias have not been analyzed sufficiently using ML methods. To portray the landscape and potential of ML research in AD and NPS studies, we present a comprehensive literature review of existing ML approaches and commonly studied AD biomarkers. We conducted PubMed searches with keywords related to NPS, AD biomarkers, machine learning, and cognition. We included a total of 38 articles in this review after excluding some irrelevant studies from the search results and including 6 articles based on a snowball search from the bibliography of the relevant studies. We found a limited number of studies focused on NPS with or without AD biomarkers. In contrast, multiple statistical machine learning and deep learning methods have been used to build predictive diagnostic models using commonly known AD biomarkers. These mainly included multiple imaging biomarkers, cognitive scores, and various omics biomarkers. Deep learning approaches that combine these biomarkers or multi-modality datasets typically outperform single-modality datasets. We conclude ML may be leveraged to untangle the complex relationships of NPS and AD biomarkers with cognition. This may potentially help to predict the progression of MCI or dementia and develop more targeted early intervention approaches based on NPS.

High-grade gliomas are the most common and aggressive adult primary brain tumors with a median survival of only 12–15 months. Current standard therapy consists of maximal safe surgical resection followed by DNA-damaging agents, such as irradiation and chemotherapy that can delay but not prevent inevitable recurrence. Some have interpreted glioma recurrence as evidence of glioma stem cells which persist in a relatively quiescent state after irradiation and chemotherapy, before the ultimate cell cycle re-entry and glioma recurrence. Conversely, latent cancer cells with a therapy-induced senescent phenotype have been shown to escape senescence, giving rise to more aggressive stem-like tumor cells than those present in the original tumor. Therefore, approaches are needed to either eliminate or keep these glioma initiating cells in a senescent state for a longer time to prolong survival. In our current study, we demonstrate that the radiation-induced cell cycle inhibitor P21 can provide a powerful route to induce cell death in short-term explants of PDXs derived from three molecularly diverse human gliomas. Additionally, cells not killed by P21 overexpression were maintained in a stable senescent state for longer than control cells. Collectively, these data suggest that P21 activation may provide an attractive therapeutic target to improve therapeutic outcomes.

Brain and spinal tumors affect 1 in 1000 people by 25 years of age, and have diverse histological, biological, anatomical and dissemination characteristics. A mortality of 30–40% means the majority are cured, although two-thirds have life-long disability, linked to accumulated brain injury that is acquired prior to diagnosis, and after surgery or chemo-radiotherapy. Only four drugs have been licensed globally for brain tumors in 40 years and only one for children. Most new cancer drugs in clinical trials do not cross the blood–brain barrier (BBB). Techniques to enhance brain tumor drug delivery are explored in this review, and cover those that augment penetration of the BBB, and those that bypass the BBB. Developing appropriate delivery techniques could improve patient outcomes by ensuring efficacious drug exposure to tumors (including those that are drug-resistant), reducing systemic toxicities and targeting leptomeningeal metastases. Together, this drug delivery strategy seeks to enhance the efficacy of new drugs and enable re-evaluation of existing drugs that might have previously failed because of inadequate delivery. A literature review of repurposed drugs is reported, and a range of preclinical brain tumor models available for translational development are explored.

563 Background: Cholangiocarcinoma is a rare malignancy with poor prognosis and outcome despite therapy. It accounts for 2.2% of all new cancer cases and 5% of all cancer deaths. Surgical resection is still the main therapeutic approach, whereas the role of peri-operative chemotherapy is debatable. Methods: The National Cancer Database (NCDB) was queried for patients diagnosed with non-metastatic biliary adenocarcinoma at age 18 or older between 2004 and 2019. After excluding patients with unknown timing of surgery and chemotherapy, patients who died within 90 days of the most definitive primary site surgery and patients lost to follow-up, we split the cohort into three groups according to the clinical stage (stage I-III). Then, we evaluated the overall survival (OS) between the different treatment modalities (surgery only, adjuvant chemotherapy and neoadjuvant chemotherapy) in each group. We studied the OS using Kaplan-Meier estimates and multivariate cox regression analyses to evaluate factors associated with OS. Results: A total of 35,260 patients with non-metastatic cholangiocarcinoma were included in the analysis, of which 50.4% were females, 83% Caucasians, 9.5% African Americans. The median age at diagnosis was 70 (range 18-90). 14,757 (41.9%) were stage I, 12,472 (35.4%) stage II and 8,031 (22.8%) stage III. 7,286 (20.7%) had surgical resection only, 8,144 (23.1%) had chemotherapy only, 6,964 (19.7%) had surgical resection with perioperative chemotherapy and 12,866 (36.5%) did not receive any treatment. We compared survival between different treatment modalities based on clinical stage. In stage I, we found patient who were treated with surgery only had better median OS (mOS) compared to adjuvant chemotherapy (65.7 vs 50.4 months, P<0.001) and no statistically significant difference between neoadjuvant chemotherapy and surgery only (mOS 79.8 vs 65.7 months, P=0.63). Whereas in stage II, patients who were treated with adjuvant and neoadjuvant chemotherapy had better mOS compared to those treated with surgery only (33.9 and 40.3 vs 29.9 months with P<0.001 and P=0.005, respectively). Same trend was seen in stage III, patients who were treated with adjuvant and neoadjuvant chemotherapy had better mOS compared to surgery only (22.6 and 41.5 vs 19.5 months, respectively with P<0.001 for all). In multivariate analysis, adjuvant and neoadjuvant chemotherapy did not affect the OS in all stages, except in stage III where neoadjuvant chemotherapy was associated with better OS (HR 0.646 95% CI 0.530-0.786; P<0.001). Conclusions: Adjuvant and neoadjuvant chemotherapy do not seem to have survival benefit in early stage (stage I and II) cholangiocarcinoma. Whereas neoadjuvant chemotherapy tends to improve OS in stage III.

AbstractData-driven machine-learning methods on neuroimaging (e.g. MRI) are of great interest for the investigation and classification of neurological diseases. However, traditional machine learning requires domain knowledge to delineate the brain regions first, followed by feature extraction from the regions. Compared with this semi-automated approach, recently developed deep learning methods have advantages since they do not require such prior knowledge; instead, deep learning methods can automatically find features that differentiate MRIs from different cohorts. In the present study, we developed a deep learning-based classification pipeline distinguishing brain MRIs of individuals with one of three types of headaches [migraine (n = 95), acute post-traumatic headache (n = 48) and persistent post-traumatic headache (n = 49)] from those of healthy controls (n = 532) and identified the brain regions that most contributed to each classification task. Our pipeline included: (i) data preprocessing; (ii) binary classification of healthy controls versus headache type using a 3D ResNet-18; and (iii) biomarker extraction from the trained 3D ResNet-18. During the classification at the second step of our pipeline, we resolved two common issues in deep learning methods, limited training data and imbalanced samples from different categories, by incorporating a large public data set and resampling among the headache cohorts. Our method achieved the following classification accuracies when tested on independent test sets: (i) migraine versus healthy controls—75% accuracy, 66.7% sensitivity and 83.3% specificity; (2) acute post-traumatic headache versus healthy controls—75% accuracy, 66.7% sensitivity and 83.3% specificity; and (3) persistent post-traumatic headache versus healthy controls—91.7% accuracy, 100% sensitivity and 83.3% specificity. The most significant biomarkers identified by the classifier for migraine were caudate, caudal anterior cingulate, superior frontal, thalamus and ventral diencephalon. For acute post-traumatic headache, lateral occipital, cuneus, lingual, pericalcarine and superior parietal regions were identified as most significant biomarkers. Finally, for persistent post-traumatic headache, the most significant biomarkers were cerebellum, middle temporal, inferior temporal, inferior parietal and superior parietal. In conclusion, our study shows that the deep learning methods can automatically detect aberrations in the brain regions associated with different headache types. It does not require any human knowledge as input which significantly reduces human effort. It uncovers the great potential of deep learning methods for classification and automatic extraction of brain imaging–based biomarkers for these headache types.

Abstract Gliomas are genomically heterogeneous tumors that may harness a convergent and therapeutically targetable set of metabolic pathways. At present, the metabolic landscape of in situ human gliomas remains incompletely characterized, hampering translational progress. To that end, we are leveraging the previously untapped potential of high molecular weight microdialysis to determine the global extracellular metabolic profiles of live human gliomas. Under an investigational device exemption, HMW microdialysis (&lt; 100 kDa) was performed at 2.0 μL/min in an initial discovery cohort of five patients in glioma and adjacent brain during neurosurgical resection; a subsequent cohort of five patients was independently analyzed to critically evaluate results from the discovery group. Untargeted metabolomics via ultra-performance liquid chromatography-tandem mass spectrometry revealed over 300 named metabolites and five drugs from only 20 μL of microdialysate, representing a short and feasible 10 minutes of intraoperative collection time. Enrichment analysis of each patient’s tumor vs. brain ranked extracellular metabolome highlighted marked metabolic convergence within the most aggressive regions of molecular diverse tumors (FDR = 0). Pathway analysis revealed significant enrichment for large neutral amino acid pathways, including valine, leucine, and isoleucine biosynthesis (p=1.6E-9) and degradation (p=0.001) and glycine, serine, and threonine metabolism (p=4.7E-5). Notably, this amino acid signature was not as abundantly present in nonenhancing tumor when compared to enhancing tumor (Average tumor/brain: 1.9x vs. 4.3x, respectively), suggesting upregulation of neutral amino acid transporters in enhancing tumor or delivery from plasma into the CNS via a disrupted BBB. Interestingly, guanidinoacetate (GAA) was our most highly conserved and upregulated metabolite (128.9x in tumor vs. brain). Given its co-production with ornithine, the precursor to protumorigenic polyamines, we posit that GAA may serve as a biomarker of increased ODC activity in live human gliomas. In conclusion, intraoperative HMW microdialysis feasibly offers improved opportunities to perform glioma metabolic biomarker and therapeutic discovery.

Abstract The relative inaccessibility of a patient’s glioma throughout their disease course precludes individualized insights into tumor biology and responses thereof to therapeutic challenge. We propose that serial cerebrospinal fluid (CSF) samples can be used to perform glioma espionage in the form of biomarker discovery for analytes that are sensitive to tumor burden (for monitoring) and candidate therapies (for biological impact). To that end, we are routinely leveraging neurosurgical access to the CNS to collect CSF samples intraoperatively and from chronically implanted Ommaya reservoirs for longitudinal CSF access (NCT04692337, NCT04692324). Global proteomic analysis of CSF was performed on the Somalogic platform – an aptamer-based technology for highly sensitive and specific analysis of over 7,000 proteins. Using each patient’s ranked protein list (glioma vs. 3 control patients), enrichment analysis was performed in a discovery cohort of five patients to identify conserved proteins of glioma. To reveal candidate glioma monitoring, further enrichment analysis was performed to identify proteins that decreased with resection and increased with recurrence, using two sets of paired patient samples. Forty-four proteins met these enrichment criteria; this candidate high-grade glioma proteomic signature was highly significantly enriched in an independent cohort of twelve glioma patients (FDR=0 for all patients). The HGG signature decreased in abundance following resection in serial samples from two other patients, confirming decreased tumor burden. Top proteins included LANC2, DPYL3, Sorcin, Integrin A1B1, and lactate dehydrogenase (FC of glioma vs. control: 145.3x, 113.9x, 78.0x, 71.6x, and 62.7x, respectively). Interestingly, despite having explicitly chosen samples with little-to-no hemoglobin in the discovery cohort, this HGG signature was highly enriched for plasma-like proteins (FDR=0), suggesting a potential biological impact of blood-brain barrier disruption on the glioma extracellular proteome. In conclusion, our data demonstrate that the glioma extracellular proteome can feasibly be accessed to perform biomarker discovery (“espionage”) via serially-acquired CSF samples.