MHC-Hammer decodes HLA disruption in tumors.

Tumors utilize various mechanisms of HLA disruption in order to evade immune surveillance. Previous computational tools have interrogated specific aspects of this process, yet a holistic picture of HLA loss of heterozygosity (LOH), transcriptomic suppression, and alternative splicing has remained challenging. In a recent Nature Genetics study, Puttick and colleagues introduced MHC Hammer, a robust computational toolkit designed to dissect the complexities of HLA disruptions that mediate immune evasion in cancer.

Midface Skeletal Sexual Dimorphism: Lessons Learned from Advanced Three-dimensional Imaging in the White Population.

Background: Facial shape is significantly influenced by the underlying facial bony skeleton. Sexual dimorphisms in these structures are crucial for craniofacial, aesthetic, and gender-affirming surgery. Previous studies have examined the orbits and upper face, but less is known about the midface.

Integrated Germline and Somatic Features Reveal Divergent Immune Pathways Driving Response to Immune Checkpoint Blockade.

Immune checkpoint blockade (ICB) has revolutionized cancer treatment; however, the mechanisms determining patient response remain poorly understood. Here, we used machine learning to predict ICB response from germline and somatic biomarkers and interpreted the learned model to uncover putative mechanisms driving superior outcomes. Patients with higher infiltration of T-follicular helper cells had responses even in the presence of defects in the MHC class-I (MHC-I).

Genetic association and machine learning improves discovery and prediction of type 1 diabetes.

Type 1 diabetes (T1D) has a large genetic component, and expanded genetic studies of T1D can lead to novel biological and therapeutic discovery and improved risk prediction. In this study, we performed genetic association and fine-mapping analyses in 817,718 European ancestry samples genome-wide and 29,746 samples at the MHC locus, which identified 165 independent risk signals for T1D of which 19 were novel. We used risk variants to train a machine learning model (named T1GRS) to predict T1D, which highly differentiated T1D from non-disease and type 2 diabetes (T2D) in Europeans as well as African Americans at or beyond the level of current standards.