Results 341 - 350 of 23740

Epidemiologic research dates back to at least the 17th century, when a haberdasher in London, John Graunt, used routinely collected data to discover that more males are born than females, and that plague epidemics are not always coincident with the coronation of a new king. In the 19th century, John Snow, considered the father of epidemiology, conducted remarkable studies that famously showed cholera to be spread via fecal-oral transmission. This research did not require computers or data processing technology. Despite the reliance of much of modern day epidemiologic research on sophisticated analytic methods and on adroit combination of electronic data sources, the conduct of epidemiologic research is not fundamentally dependent on technology. Epidemiologists have of course capitalized on the use of digital technology and data storehouses to circumvent the difficulty of collecting vast amounts of data for each new research study. But technology brings with it new problems, which I intend to illustrate.

Abstract not available.

Healthcare systems are experiencing rising demand alongside a rapid expansion of medical knowledge, creating challenges in delivering accurate, up-to-date treatment for individual patients and in advancing proactive and preventive care. This talk will explore how medical data can be leveraged to support predictive, proactive, and personalized care, and will outline the development of various medical AI products and their integration into point-of-care workflows.

The talk will also focus on the challenges of practicing responsible AI in healthcare, including ensuring safety and fairness when AI systems are deployed in new clinical settings where performance may change in unexpected ways. It will highlight the critical role of healthcare providers in evaluating, implementing, and continuously monitoring these technologies to ensure reliable performance and maintain high-quality, trustworthy care.

Cancer genomics has traditionally focused on identifying risk alleles and driver mutations through statistical association, often overlooking the functional and evolutionary principles shaping cancer predisposition and progression. Here, we revisit cancer genetics through a functionalist and evolutionary lens, integrating germline and somatic variation to uncover both expected and unexpected genetic forces acting in cancer.

We first analyze cancer predisposition across ten major cancer types in the UK Biobank (UKB) using proteome-wide association studies (PWAS), a gene-based framework that links inherited variation to predicted protein functional damage. This approach identifies 110 significant gene–cancer associations, nearly half of which exhibit an unexpected protective effect, whereby damaging variants are associated with reduced cancer risk. Moreover, PWAS reveals that 46% of significant associations act exclusively through recessive inheritance, highlighting a substantial class of predisposition effects largely missed by additive GWAS models. Combined with classical GWAS, we identify 145 cancer-associated loci, including 51 previously unreported regions, underscoring the diversity of genetic architectures underlying cancer risk.

Focusing on breast cancer, we integrate population-scale PWAS with family-based exome sequencing, revealing both known and novel predisposition genes. These findings emphasize the context-specific and lineage-dependent nature of inherited cancer susceptibility. To bridge germline predisposition with tumor evolution, we introduce FABRIC, a functionalist framework that quantifies gene-level selection in cancer by comparing observed somatic mutations against gene-specific functional backgrounds. Applying FABRIC to over 10,000 tumors uncovers approximately 600 genes under significant negative or positive selection, including over 180 coding genes previously overlooked. Together, these results argue for a revised view of cancer genetics, one that embraces functional impact, recessive inheritance, protective effects, and evolutionary selection. This integrative framework expands the catalog of cancer-relevant genes and offers new avenues for early diagnosis, genetic counseling, and personalized cancer risk assessment. We will also share some ongoing research on antigen-presenting machinery (APM) where alternative evolutionary routes dictate success or failure in modern cancer immunotherapy.

To read more see: (i) Brandes, N. et al. (2021) Scientific reports 11, 14901; (ii) Kelman et al (2021) Cancer Research 81, 1178; (iii) Passi, G., et al. (2024) Briefings in Bioinformatics 25 (4), bbae346. (iv) Zok and Linial (2025) BioRxiv doi:10.1101/2025.09.22.677872.

In this talk, we'll explore the perspective that the computer science literature gives on research on sensitive data---what can be done safely, and also how privacy concerns can get us into trouble, sometimes in surprising ways.