This presentation highlights VarSeq’s advanced genomic analysis capabilities, including precise CNV calling for whole exome sequencing, long-read support with allele phasing, and VSPGx for pharmacogenomics, adhering to CPIC and FDA standards. We will discuss benchmark data, case studies, and VarSeq’s “Bring Your Own Cloud” flexibility for deployment on AWS, Azure, Google Cloud, or on-premises, making it a versatile tool for precision medicine.

This talk presents ongoing efforts to extensively characterize the first tumor-normal benchmark genomes from the Genome in a Bottle consortium. A tumor cell line was established from a pancreatic ductal adenocarcinoma sample that was consented for the public release of genomic data. This dataset currently includes WGS measurements from Illumina, ONT, PacBio HiFi, Hi-C, Bionano, Bioskryb single cell, PacBio Onso, Element, and Ultima as well as karyotype data. An open working group of the GIAB consortium is developing somatic variant benchmarks from this data.

Computational analysis of large-scale metagenomics sequencing datasets have proven to be both incredibly valuable for extracting isolate-level taxonomic, and functional insights from complex microbial communities. However, due to an ever-expanding ecosystem of metagenomics-specific methods and file-formats, designing studies which implement seamless and scalable end-to-end workflows, and exploring the massive amounts of output data have become studies unto themselves. To address the gap in easily extensible yet robustly distributable metagenomics workflows, we have developed a module-based metagenomics analysis system: "Core Analysis Metagenomics Pipeline" (CAMP) along with a standardized module and working directory architecture. 

New “hybrid” secondary analysis pipeline combining 30x short-read sequencing and 10x PacBio HiFi sequencing resulting in possibly the most accurate genome analysis available. Benchmarking with a 30x NovaSeq and 10x HiFi coverage, we record only 5,909/1,665 total (SNP/indel) errors relative to the GIAB v4.2.1 HG002 truthset. The pipeline reduces variant calling errors by 74% compared to DeepVariant on 30x NovaSeq data and 26% relative to DV with 30x HiFi data.

This talk delves into the integration of spatial transcriptomics and single-cell multiomics with digital pathology using advanced AI techniques, including vision transformers, to revolutionize data interpretation. By imputing hidden features and patterns from spatial transcriptomics, the presentation will showcase groundbreaking insights into cellular microenvironments and disease pathology, offering unprecedented resolution for precision medicine. Attendees will explore how combining these innovations accelerates advancements in oncology and health research.

In osteoporosis research, the analysis of osteoclast cell images is critical but often burdensome, relying heavily on manual processing and expert annotations. We introduce a pioneering machine learning algorithm that automates this analysis, leveraging weak supervision to incorporate nuclear data while addressing domain shift challenges in transferring knowledge from the data-rich mouse domain to the critical human context. Our work exemplifies how advanced AI and computer vision techniques can tackle significant issues in translational osteoporosis research, highlighting the transformative role of AI in drug discovery and development. Initial findings were presented at a workshop at the prestigious 2024 Computer Vision & Pattern Recognition (CVPR) conference. This talk will provide insights into the application of cutting-edge computer vision tools in translational health research and offer strategies for successful AI-biomedical collaborations, underscoring AI's potential to enhance the efficiency and efficacy of therapeutic development.

Modern AI methods are good at grading cancer, but they are not used much in hospitals yet. We developed a new online tool that combines human expertise and smart AI predictions to grade prostate cancer. Experienced doctors helped us improve the tool by answering questions in a survey and a test called the NASA TLX Usability Test. This helped ensure that the tool meets the needs of intended medical users and proves a valuable addition in clinics. The feedback obtained was positive for three themes that included acceptability, ease of use, and understanding with suggested application feature improvements as feedback for real-world hospital use. The task completion process was not very demanding overall. The web application has potential to serve well upon minor improvements based on feedback.