From paper to excel sheets, electronic lab notebooks, and finally to the cloud, biomedical data has been going places in the last few decades! As we transition into a post-cloud world and all your data is on the cloud, any AI initiative will be at risk unless the data is clean and linked to each other. In this session, Abhishek Jha elaborates on how Elucidata is doubling and tripling down on a data-centric approach to accelerate biological discovery.

Published clinical trial data holds rich information on treatments and clinical endpoints that can be useful for future trial design and clinical decisions. This data is usually locked in text form, and not easily accessible. Making such published clinical trial data FAIR is challenging. Even with such data being accessible there remain added complexities in using the data for mathematical models. This talk will explore the various stages of compiling useful datasets from published data all the way to leveraging this clinical knowledge to generate insight and enable data-informed decisions. Challenges and opportunities in this data journey will be discussed.

As the world is moving at breakneck speed toward larger models to improve accuracy, we take a step back from the rat race to take a more efficient and realistic look at the technology and explore how it can be improved in a more optimized and well-rounded manner. Join the exciting conversation between ML/NLP expert Dr. Ashutosh Modi and our BioNLP Director, Shashank Jatav, to learn more.

As modeling projects grow, so grow the costs of debugging, scaling, and modifying the model pipeline. One method to minimize the costs of model maintenance is to train models in reproducible iterations. In the context of machine learning, we define a reproducible model iteration as the output of an executable script that is a pure function of three variables: code, environment, and data. Reproducible models are not an end, but a means to faster, more correct iterations. A reproducible model history implies that developers can confidently reconstruct any past model iteration. As a result, reproducibility makes it easier for developers to experiment with modifications, isolate bugs, and revert to known good iterations when problems arise.

The presentation will use metabolomics as a data construct to highlight the importance of standardization of both protocols and data to enable data-driven analytics for the creation of actionable solutions. A case study will be presented on the use of metabolomics data for target identification and validation of its potential clinical utility as a pathway for drug discovery.

Predictive models used in drug discovery require a viable level of data quality. A faulty model can lead to completely off-the-mark predictions and sunk project costs.In sharp contrast, much of the available biomedical data is unstructured and prone to errors due to varying experimental protocols (incomplete metadata information, missing annotations, inconsistent file formats). To ensure their datasets are ML-Ready, R&D teams must set up a system that continuously assesses and iterates on the data and metadata quality. This session will demonstrate Elucidata’s data quality assessment approach, which ensures an input dataset is standardized and has accurate, complete, and a breadth of metadata information before it is considered model quality.

Today, biomolecular research produces a large amount of data. This data provides important clues for discovery, translational research, and personalized medicine. At Elucidata, our aim is to build a platform that puts high-quality data central to ML workflows in pre-clinical R&D.  In this session, our speakers shed light on the Do-it-yourself (DIY) data journey on Polly, starting from ingesting raw data on Polly's cloud, analyzing and visualizing processed data and everything in between. Join Neychelle, Director - Solutions and Technical Sales, and Jainik Dedhia, Senior product manager, as they give us insights into what’s coming up next quarter on Polly.