Benchmark of Data Processing Methods and Machine Learning Models for Gut Microbiome-Based Diagnosis of Inflammatory Bowel Disease.

Patients with inflammatory bowel disease (IBD) wait months and undergo numerous invasive procedures between the initial appearance of symptoms and receiving a diagnosis. In order to reduce time until diagnosis and improve patient wellbeing, machine learning algorithms capable of diagnosing IBD from the gut microbiome's composition are currently being explored. To date, these models have had limited clinical application due to decreased performance when applied to a new cohort of patient samples. Various methods have been developed to analyze microbiome data which may improve the generalizability of machine learning IBD diagnostic tests. With an abundance of methods, there is a need to benchmark the performance and generalizability of various machine learning pipelines (from data processing to training a machine learning model) for microbiome-based IBD diagnostic tools. We collected fifteen 16S rRNA microbiome datasets (7,707 samples) from North America to benchmark combinations of gut microbiome features, data normalization and transformation methods, batch effect correction methods, and machine learning models. Pipeline generalizability to new cohorts of patients was evaluated with two binary classification metrics following leave-one-dataset-out cross (LODO) validation, where all samples from one study were left out of the training set and tested upon. We demonstrate that taxonomic features processed with a compositional transformation method and batch effect correction with the naive zero-centering method attain the best classification performance. In addition, machine learning models that identify non-linear decision boundaries between labels are more generalizable than those that are linearly constrained. Lastly, we illustrate the importance of generating a curated training dataset to ensure similar performance across patient demographics. These findings will help improve the generalizability of machine learning models as we move towards non-invasive diagnostic and disease management tools for patients with IBD.

Role of Digital Health and Artificial Intelligence in Inflammatory Bowel Disease: A Scoping Review.

Inflammatory bowel diseases (IBD), subdivided into Crohn's disease (CD) and ulcerative colitis (UC), are chronic diseases that are characterized by relapsing and remitting periods of inflammation in the gastrointestinal tract. In recent years, the amount of research surrounding digital health (DH) and artificial intelligence (AI) has increased. The purpose of this scoping review is to explore this growing field of research to summarize the role of DH and AI in the diagnosis, treatment, monitoring and prognosis of IBD. A review of 21 articles revealed the impact of both AI algorithms and DH technologies; AI algorithms can improve diagnostic accuracy, assess disease activity, and predict treatment response based on data modalities such as endoscopic imaging and genetic data. In terms of DH, patients utilizing DH platforms experienced improvements in quality of life, disease literacy, treatment adherence, and medication management. In addition, DH methods can reduce the need for in-person appointments, decreasing the use of healthcare resources without compromising the standard of care. These articles demonstrate preliminary evidence of the potential of DH and AI for improving the management of IBD. However, the majority of these studies were performed in a regulated clinical environment. Therefore, further validation of these results in a real-world environment is required to assess the efficacy of these methods in the general IBD population.

Development of a generalizable natural language processing pipeline to extract physician-reported pain from clinical reports: Generated using publicly-available datasets and tested on institutional clinical reports for cancer patients with bone metastases.

OBJECTIVE: The majority of cancer patients suffer from severe pain at the advanced stage of their illness. In most cases, cancer pain is underestimated by clinical staff and is not properly managed until it reaches a critical stage. Therefore, detecting and addressing cancer pain early can potentially improve the quality of life of cancer patients. The objective of this research project was to develop a generalizable Natural Language Processing (NLP) pipeline to find and classify physician-reported pain in the radiation oncology consultation notes of cancer patients with bone metastases. MATERIALS AND METHODS: The texts of 1249 publicly-available hospital discharge notes in the i2b2 database were used as a training and validation set. The MetaMap and NegEx algorithms were implemented for medical terms extraction. Sets of NLP rules were developed to score pain terms in each note. By averaging pain scores, each note was assigned to one of the three verbally-declared pain (VDP) labels, including no pain, pain, and no mention of pain. Without further training, the generalizability of our pipeline in scoring individual pain terms was tested independently using 30 hospital discharge notes from the MIMIC-III database and 30 consultation notes of cancer patients with bone metastasis from our institution's radiation oncology electronic health record. Finally, 150 notes from our institution were used to assess the pipeline's performance at assigning VDP. RESULTS: Our NLP pipeline successfully detected and quantified pain in the i2b2 summary notes with 93% overall precision and 92% overall recall. Testing on the MIMIC-III database achieved precision and recall of 91% and 86% respectively. The pipeline successfully detected pain with 89% precision and 82% recall on our institutional radiation oncology corpus. Finally, our pipeline assigned a VDP to each note in our institutional corpus with 84% and 82% precision and recall, respectively. CONCLUSION: Our NLP pipeline enables the detection and classification of physician-reported pain in our radiation oncology corpus. This portable and ready-to-use pipeline can be used to automatically extract and classify physician-reported pain from clinical notes where the pain is not otherwise documented through structured data entry.

Role of artificial intelligence in the care of patients with nonsmall cell lung cancer.

BACKGROUND: Lung cancer is the leading cause of cancer death worldwide. In up to 57% of patients, it is diagnosed at an advanced stage and the 5-year survival rate ranges between 10%-16%. There has been a significant amount of research using machine learning to generate tools using patient data to improve outcomes. METHODS: This narrative review is based on research material obtained from PubMed up to Nov 2017. The search terms include "artificial intelligence," "machine learning," "lung cancer," "Nonsmall Cell Lung Cancer (NSCLC)," "diagnosis" and "treatment." RESULTS: Recent studies support the use of computer-aided systems and the use of radiomic features to help diagnose lung cancer earlier. Other studies have looked at machine learning (ML) methods that offer prognostic tools to doctors and help them in choosing personalized treatment options for their patients based on molecular, genetics and histological features. Combining artificial intelligence approaches into health care may serve as a beneficial tool for patients with NSCLC, and this review outlines these benefits and current shortcomings throughout the continuum of care. CONCLUSION: We present a review of the various applications of ML methods in NSCLC as it relates to improving diagnosis, treatment and outcomes.

Pattern of Local Recurrence and Distant Metastasis in Breast Cancer By Molecular Subtype.

BACKGROUND AND PURPOSE:  No longer considered a single disease entity, breast cancer is being classified into several distinct molecular subtypes based on gene expression profiling. These subtypes appear to carry prognostic implications and have the potential to be incorporated into treatment decisions. In this study, we evaluated patterns of local recurrence (LR), distant metastasis (DM), and association of survival with molecular subtype in breast cancer patients in the post-adjuvant radiotherapy setting. MATERIAL AND METHODS:  The medical records of 1,088 consecutive, non-metastatic breast cancer patients treated at a single institution between 2004 and 2012 were reviewed. Estrogen/progesterone receptors (ER/PR) and human epidermal growth factor receptor-2 (HER2) enrichment were evaluated by immunohistochemistry. Patients were categorized into one of four subtypes: luminal-A (LA; ER/PR+, HER2-, Grade 1-2), luminal-B (LB; ER/PR+, HER2-, Grade > 2), HER2 over-expression (HER2; ER/PR-, HER2+), and triple negative (TN; ER/PR-, HER2-).  Results: The median follow-up time was 6.9 years. During the follow-up, 16% (174/1,088) of patients failed initial treatment and developed either LR (48) or DM (126). The prevalence of LR was the highest in TN (12%) and the lowest in LA (2%). Breast or chest wall relapse was the most frequent site (≈80%) of recurrence in LA, LB, and HER2 subtypes, whereas the regional lymph nodes and chest wall were the common sites of relapse in the TN group (50.0%). DM rates were 6.4% in LA, 12.1% in LB, 19.2% in HER2, and 27.4% in TN subgroups. Five-year survival rates were 84%, 83%, 84%, and 77% in the LA, LB, HER2 and TN subgroups, respectively. There was a statistically significant association between survival and molecular subtypes in an univariate analysis. In the adjusted multivariate analysis, the following variables were independent prognostic factors for survival: T stage, N stage, and molecular subtype. CONCLUSIONS:  Of the four subtypes, the LA subtype tends to have the best prognosis, fairly high survival, and low recurrent or metastases rates. The TN and HER2 subtypes of breast cancer were associated with significantly poorer overall survival and prone to earlier recurrence and metastases. Our results demonstrate a significant association between molecular subtype and survival. The risk of death and relapse/metastases increases fewfold in TN compared to LA. Future prospective studies are warranted and could ultimately lead to the tailoring of adjuvant radiotherapy treatment fields based on both molecular subtype and the more conventional clinicopathologic characteristics.

Combination of cyclophosphamide, rituximab, and intratumoral CpG oligodeoxynucleotide successfully eradicates established B cell lymphoma.

Rituximab plus chemotherapy is standard therapy for patients with non-Hodgkin B cell lymphoma, but often complete response or cure is not achieved. Toll-like receptor 9 agonist CpG oligodeoxynucleotides (CpG) can improve antibody-dependent cellular cytotoxicity and adaptive antitumor immune responses. Using a syngeneic murine B cell lymphoma expressing human CD20 (38C13-huCD20), we previously demonstrated that rituximab plus intratumoral CpG, but not systemic CpG, could eradicate up to half of 7-day established 38C13-huCD20 tumors. However, larger 10-day established tumors could not be cured with this regimen. We thus hypothesized that cytoreduction with cyclophosphamide (Cy) before immunotherapy might permit eradication of these more advanced tumor burdens. Pretreatment with Cy resulted in tumor eradication from 83% of animals treated with rituximab/CpG, whereas Cy/CpG or Cy/rituximab treatments only cured 30% or 17%, respectively (P<0.005). Tumor eradication depended on natural killer cells, but not T cells, macrophages, or complement. Only mice treated with Cy/rituximab/CpG partially resisted rechallenge with tumor cells. Foxp3 Treg and CD11bGr1 myeloid suppressor cells persisted within lymphoid organs after therapy, possibly influencing the ability to establish adaptive tumor immunity. In conclusion, cytoreduction with Cy permitted the cure of large, established lymphomas not otherwise responsive to rituximab plus intratumoral CpG immunotherapy.

Intratumoral but not systemic delivery of CpG oligodeoxynucleotide augments the efficacy of anti-CD20 monoclonal antibody therapy against B cell lymphoma.

The anti-CD20 monoclonal antibody rituximab (Rituxan) has become a mainstay in the treatment of B cell non-Hodgkin lymphomas. The mechanisms of action for rituximab include antibody-dependent cellular cytotoxicity (ADCC), complement-dependent cytotoxicity, and apoptosis induction. Combination of anti-CD20 antibodies with immunostimulatory agents may improve their efficacy via enhancement of one or more of these mechanisms. Toll-like receptor 9 agonist CpG oligodeoxynucleotides administered systemically have been studied in clinical trials with and without rituximab. However, recent data suggest that intratumoral (IT) delivery of CpG has advantages in the treatment of tumors. Using a syngeneic murine B cell lymphoma line expressing human CD20, we found that IT, but not systemically administered CpG significantly improved the efficacy of rituximab against 7-day established tumors. Rituximab plus IT CpG could eradicate tumors from 42% of mice, whereas systemically administered CpG, with or without rituximab, did not achieve tumor eradication. Both natural killer cells and complement participated in the cure of tumors by rituximab plus IT CpG, apparently by increasing tumor cell sensitivity to complement and ADCC lysis, and by augmenting the cytotoxicity of ADCC effectors. No role for T cells in mediating tumor eradication was demonstrated in this model. These results suggest that previous clinical trials in B cell lymphoma combining systemic administration of CpG with rituximab may have employed suboptimal routes of CpG delivery. Future trials combining IT CpG with anti-CD20 antibodies or the antibody-mediated targeting of CpG directly to the sites of B cell lymphoma may thus be warranted.

Enhanced immune stimulation by a therapeutic lymphoma tumor antigen vaccine produced in insect cells involves mannose receptor targeting to antigen presenting cells.

Therapeutic vaccination of lymphoma patients with tumor-specific immunoglobulin (idiotype, Id) coupled to the carrier protein keyhole limpet hemocyanin (Id-KLH) is undergoing clinical investigation, and methods to improve the immunogenicity of these and other protein tumor antigen vaccines are being sought. Id proteins can be produced via tumor-myeloma hybridomas or recombinant methods in mammalian, bacteria, or insect cells. We now demonstrate that terminal mannose residues, characteristic of recombinant proteins produced in insect cells, yield Id proteins with significantly enhanced immunostimulatory properties compared to Id proteins derived from mammalian cells. Recombinant baculovirus-infected insect cell-derived Id showed higher binding to and activation of human dendritic cells mediated by mannose receptors. In vivo, insect cell-derived Id elicited higher levels of tumor-specific CD8+ cytotoxic T lymphocyte (CTL) and improved eradication of pre-established murine lymphoma. Insect cell and mammalian Id generated similar levels of tumor-specific antibodies, showing no impairment in antibody responses to native tumor antigen despite the glycoslylation differences in the immunogen. Combining insect cell production and maleimide-based KLH conjugation offered the highest levels of anti-tumor immunity. Our data comparing sources of recombinant Id protein tumor antigens used in therapeutic cancer vaccines demonstrate that insect cell-derived antigens can offer several immunologic advantages over proteins derived from mammalian sources.

Maleimide conjugation markedly enhances the immunogenicity of both human and murine idiotype-KLH vaccines.

The collection of epitopes present within the variable regions of the tumor-specific clonal immunoglobulin expressed by B cell lymphomas (idiotype, Id) can serve as a target for active immunotherapy. Traditionally, tumor-derived Id protein is chemically conjugated to the immunogenic foreign carrier protein keyhole limpet hemocyanin (KLH) using glutaraldehyde to serve as a therapeutic vaccine. While this approach offered promising results for some patients treated in early clinical trials, glutaraldehyde Id-KLH vaccines have failed to induce immune and clinical responses in many vaccinated subjects. We recently described an alternative conjugation method employing maleimide-sulfhydryl chemistry that significantly increased the therapeutic efficacy of Id-KLH vaccines in three different murine B cell lymphoma models, with protection mediated by either CD8(+) T cells or antibodies. We now define in detail the methods and parameters critical for enhancing the in vivo immunogenicity of human as well as murine Id-KLH conjugate vaccines. Optimal conditions for Id sulfhydryl pre-reduction were determined, and maleimide Id-KLH conjugates maintained stability and potency even after prolonged storage. Field flow fractionation analysis of Id-KLH particle size revealed that maleimide conjugates were far more uniform in size than glutaraldehyde conjugates. Under increasingly stringent conditions, maleimide Id-KLH vaccines maintained superior efficacy over glutaraldehyde Id-KLH in treating established, disseminated murine lymphoma. More importantly, human maleimide Id-KLH conjugates were consistently superior to glutaraldehyde Id-KLH conjugates in inducing Id-specific antibody and T cell responses. The described methods should be easily adaptable to the production of clinical grade vaccines for human trials in B cell malignancies.

Sulfhydryl-based tumor antigen-carrier protein conjugates stimulate superior antitumor immunity against B cell lymphomas.

Therapeutic vaccination of B cell lymphoma patients with tumor-specific Ig (idiotype, or Id) chemically coupled to the immunogenic foreign carrier protein keyhole limpet hemocyanin (KLH) using glutaraldehyde has shown promising results in early clinical trials, and phase III trials are underway. However, glutaraldehyde Id-KLH vaccines fail to elicit anti-Id immune and clinical responses in many patients, possibly because glutaraldehyde reacts with lysine, cysteine, tyrosine, and histidine residues, damaging critical immunogenic epitopes. A sulfhydryl-based tumor Ag-carrier protein conjugation system using maleimide chemistry was used to enhance the efficacy of Id-KLH vaccines. Maleimide Id-KLH conjugates eradicated A20 lymphoma from most tumor-bearing mice, whereas glutaraldehyde Id-KLH had little efficacy. Maleimide Id-KLH elicited tumor-specific IgG Abs and T cells, with CD8(+) T cells being the major effectors of antilymphoma immunity. Maleimide Id-KLH vaccines also demonstrated superior efficacy in 38C13 and BCL-1 lymphoma models, where Abs were shown to be critical for protection. Importantly, standard glutaraldehyde Id-KLH conjugation procedures could result in "overconjugation" of the tumor Ag, leading to decreased efficacy, whereas the heterobifunctional maleimide-based conjugation yielded potent vaccine product regardless of conjugation duration. Under lysosomal processing conditions, the Id-carrier protein linkage was cleavable only after maleimide conjugation. Maleimide KLH conjugation was easily performed with human Igs analogous to those used in Id-KLH clinical trials. These data support the evaluation of sulfhydryl-based Id-KLH vaccines in lymphoma clinical trials and possibly the use of tumor Ag-carrier protein vaccines for other cancers.

Dendritic cells loaded with apoptotic antibody-coated tumor cells provide protective immunity against B-cell lymphoma in vivo.

The in vitro priming of tumor-specific T cells by dendritic cells (DCs) phagocytosing killed tumor cells can be augmented in the presence of antitumor monoclonal antibody (mAb). We investigated whether DCs phagocytosing killed lymphoma cells coated with tumor-specific antibody could elicit antitumor immunity in vivo. Irradiated murine 38C13 lymphoma cells were cocultured with bone marrow-derived DCs in the presence or absence of tumor-specific mAb. Mice vaccinated with DCs cocultured with mAb-coated tumor cells were protected from tumor challenge (60% long-term survival), whereas DCs loaded with tumor cells alone were much less effective. The opsonized whole tumor cell-DC vaccine elicited significantly better tumor protection than a traditional lymphoma idiotype (Id) protein vaccine, and in combination with chemotherapy could eradicate preexisting tumor. Moreover, the DC vaccine protected animals from both wild-type and Id-negative variant tumor cells, indicating that Id is not a major target of the induced tumor immunity. Protection was critically dependent upon CD8(+) T cells, with lesser contribution by CD4(+) T cells. Importantly, opsonized whole tumor cell-DC vaccination did not result in tissue-specific autoimmunity. Since opsonized whole tumor cell-DC and Id vaccines appear to target distinct tumor antigens, optimal antilymphoma immunity might be achieved by combining these approaches.