Single-cell analysis helps biologists and bioinformaticians reveal complex and rare cell populations, uncover regulatory relationships among genes, analyze and visualize gene expression differences among cell types. In this talk we will explore new tools for analyzing, interpreting and exploring scRNA-seq data and the underlying biology. We will also show how to integrate ‘omics datasets from different platforms to gain insights into the biology and molecular drivers of specific cell populations.

Join us as we explore how modern BI drives business transformation. From data centers to deserts, we’ll discuss what’s next for business intelligence. Then we’ll highlight how modern BI is helping teams and share how to act on the three core elements of modern BI: data agility, data literacy and intelligent action. You’ll learn ways to unlock the value of data throughout your business, and drive the transformation your business needs.

In scenarios where clinical trials may not produce sufficient data on their own, such as in rare or serious conditions, historical clinical trial data can provide scientifically-rigorous evidence to fill data gaps. This session will cover the value of external data to your clinical development program, including examples from work with Cytokine Release Syndrome treated with CAR-T therapies where past trial data can help fill the gaps to:

• Understand disease risk within specific patient groups
• Analyze experimental treatments versus standard of care outcomes 
• Generate evidence to inform product development and medical engagement

The convergence of several technology trends has accelerated progress in intelligent applications. The volume of data continues to double every couple of years. AI engineers now have massive compute power they can tap into, and they are devising ever more novel algorithms. How does an enterprise approach the key technology building blocks to enable scalable and agile applications which leverage embedded machine learning, including deep learning? The talk will describe some of the foundational technologies, structure and processes that could enable an enterprise to put AI to work for transforming their businesses.

More and more machine learning algorithms in pharma research are transitioning from a one-off scenario, where the model is built and ran few times, to repeated usage of the same model in daily research workflows. This shift significantly raises the bar on the quality and setup necessary to train and deploy ML models. With an ever increasing number of models we learned that leveraging an ML platform for operationalization has scale benefits.

Gene regulatory network inference is instrumental to the discovery of genetic mechanisms driving diverse diseases, including cancer. We cast this problem as graph matching and leverage its connection to machine learning to improve the state of the art in predicting the binding of transcription factors to promoter regions of genes.

The Cambridge Structural Database (CSD) contains >1 million experimentally determined, expertly curated, small-molecule crystal structures. These structures have been collected from >400,000 authors over the last 55 years, resulting in a diverse dataset. These features make the CSD attractive for machine learning methods which benefit from large datasets. This talk will highlight some ways our researchers have utilized this structural data to gain insights in chemistry, biochemistry, and materials science.

In this study, we focused on patients diagnosed with psoriatic arthritis, an auto-immune disease that impairs the joint and bone function. In many cases, these patients were already suffering from a related auto-immune disease called psoriasis, which mainly impacts the skin, prior to developing a more debilitating psoriatic arthritis. We will present how a machine learning pipeline was developed to identify different progression pathways from a skin disease to a joint disease, and the potential impact of the methodology to inform clinical trial feasibility assessment for pharmaceutical R&D efforts, enhancing clinical trial efficiency and improving patient outcomes.

Wild-type transthyretin amyloid cardiomyopathy is a progressive, life-threatening, increasingly recognized but underdiagnosed cause of heart failure. Here we show that a random forest machine learning model can identify potential wild-type transthyretin amyloid cardiomyopathy using medical claims data. We show that the machine learning model performs well in identifying patients thereby providing a systematic framework to increase the suspicion of transthyretin cardiac amyloidosis in patients with heart failure.

In order to identify patients with cognitive concerns in electronic medical records, we applied a deep learning based natural language processing (NLP) algorithm to unstructured clinician notes and compared the model’s performance to a baseline model that used regularized logistic regression with structured data (diagnosis codes and medication data). The deep learning model improved the AUROC from 0.79 to 0.90 and increased sensitivity of dementia detection from 0.59 to 0.79. 

AILANI is a semantic search enterprise solution for fast, easy and comprehensive knowledge discovery. It combines semantic modelling, ontologies, linguistics and AI algorithms to identify relevant information. Using ontology-based refiners enables fast and efficient retrieval of information about the clinical competitive landscape as well as identification and mapping of KOLs. By integrating organization-specific content AILANI leverages knowledge buried both in decade old data and data from news feeds and clinical trials.

The life science industry needs more than an ELN - customers can use Benchling to model their scientific workflows and then analyze their data. Join this session to learn how this is currently done, advantages and disadvantages to certain modeling approaches, and analytical methods on top of that data. Session will include a live demonstration with notional data and usage of Benchling Insights and the Benchling developer platform.

We applied unsupervised machine learning on time series data collected from a clinical trial to uncover distinct patterns of patient treatment response. This talk will cover the challenges of using unsupervised machine learning on clinical trial data and the technical solutions to overcome these challenges, including data imputation, cluster optimization, secondary analysis, and clinical interpretation of results.

Precision oncology still hinges on the application of cancer therapies that are designed for the ‘average patient’ as a ‘one size fits all’ approach. Since there is no ‘average patient’, targeted treatments are successful only in some patients. For PO to be effective, technologies are needed that will allow the rapid identification of key altered pathways in each patient’s tumor that are susceptible to molecularly targeted or immunological therapies and the presentation of these in a context-sensitive fashion at the point of clinical decision-making. Virtual MTBs (VMTBs) offer a solution to this issue to help drive cancer clinical trial recruitment.

The System for High-Intensity Evaluation During Radiation Therapy (SHIELD-RT) study (ClinicalTrials.gov NCT04277650; https://ascopubs.org/doi/full/10.1200/JCO.20.01688) was a randomized controlled study that demonstrated that machine learning based on electronic health records can be implemented in the clinical setting to direct supplemental clinical evaluations during outpatient cancer radiotherapy and chemoradiation. This reduced acute care (emergency visits and hospitalization) in high risk patients from 22.3% to 12.3%. Routine implementation of this system at the Duke Cancer Institute is underway.