RESUMO
Although machine learning (ML) has shown promise across disciplines, out-of-sample generalizability is concerning. This is currently addressed by sharing multi-site data, but such centralization is challenging/infeasible to scale due to various limitations. Federated ML (FL) provides an alternative paradigm for accurate and generalizable ML, by only sharing numerical model updates. Here we present the largest FL study to-date, involving data from 71 sites across 6 continents, to generate an automatic tumor boundary detector for the rare disease of glioblastoma, reporting the largest such dataset in the literature (n = 6, 314). We demonstrate a 33% delineation improvement for the surgically targetable tumor, and 23% for the complete tumor extent, over a publicly trained model. We anticipate our study to: 1) enable more healthcare studies informed by large diverse data, ensuring meaningful results for rare diseases and underrepresented populations, 2) facilitate further analyses for glioblastoma by releasing our consensus model, and 3) demonstrate the FL effectiveness at such scale and task-complexity as a paradigm shift for multi-site collaborations, alleviating the need for data-sharing.
Assuntos
Big Data , Glioblastoma , Humanos , Aprendizado de Máquina , Doenças Raras , Disseminação de InformaçãoRESUMO
Heterochromatin differs from euchromatin by a set of specific features. We suggested earlier that specific features of heterochromatin result from differences in DNA topology of these two chromatin types and provided explanations for the majority of them (Gruzdev 2000). We proposed that, unlike topologically closed euchromatic DNA, the DNA of heterochromatin is topologically open, i.e. it likely contains single- or doublestrand breaks. In this work, we studied the topological state of DNA in a block of centromeric heterochromatin and in a euchromatic banded region of Chironomus melanotus polytene chromosomes by microfluorimetric methods using the fluorescent intercalating dye ethidium bromide (EB). It was demonstrated that the fraction of topologically closed DNA in heterochromatin blocks is five-fold smaller than in the banded region. The data obtained support the hypothesis proposed.