The Xerna TME Panel is a machine learning-based transcriptomic biomarker retrospectively validated for use in a dozen cancer types across 7 drugs of distinct target/mechanism. Developed using RNA-seq and an artificial neural network, the panel classifies tumors into four biologically defined subtypes to predict responses to immunotherapies and anti-angiogenic agents across various cancers. Central to its success was the focus on explainability, demonstrating to the FDA that the model performs as intended and ensuring clinical relevance. This talk will explore the development, validation, and regulatory journey of the Xerna TME Panel, emphasizing the critical role of explainable AI in achieving regulatory acceptance and advancing precision medicine.

This presentation explores explainable and transparent AI in healthcare and biomedical research. It introduces key dimensions addressing these challenges and outlines two approaches with application examples: 1. A network-based approach for multi-omics data integration, which is exemplified through a drug response prediction case study; 2. Generative AI for diagnostic knowledge curation, which is based on a workflow for evidence extraction and genomic knowledge updating from vast scientific literature. The talk will emphasise the multifaceted nature of explainability and transparency, highlighting their synergistic role in enhancing AI trustworthiness in biomedicine.

I will talk about why do we need explainability of AI algorithms in healthcare and give an example of use case in Tanzania.

In this presentation, I will discuss how we used miRNA expression data and AI to develop a prognosis model for ovarian cancer patients. I will also highlight how explainable AI provided valuable insights into the underlying biology of miRNAs in this context.

As a large language model, Moremi AI is able to assist in many capabilities in various disciplines. It would serve as an invaluable tool for researchers and scientists in the fields of biology, biochemistry, and drug discovery. Its applications span various areas within these fields, from understanding fundamental biological processes to facilitating the discovery and development of new drugs. In this talk, we provide a broad overview of the areas in biology, biochemistry, and drug discovery where Moremi AI can be applied based on its various biomedical tasks.

The talk will cover two types of applied AI problems: pest detection in farms, and screening for Tuberculosis. These are addressed using practical, multimodal machine learning based solutions that we develop for use in rural communities in the global South. They are designed from the ground up to work in resource-constrained settings, with data that is gathered from the field. The aim of this talk is to convey to the audience the scale and the pressing nature of the problems themselves in terms of their impact on lives and livelihoods, the resource constraints, the data challenges and technical innovations, and finally deployment pathways to ensure that they will be adopted at scale.