The use of self-supervised learning to train pathology foundation models has increased substantially in the past few years. Notably, several models trained on large quantities of clinical data have been made publicly available in recent months. This will significantly enhance scientific research in computational pathology and help bridge the gap between research and clinical deployment. With the increase in availability of public foundation models of different sizes, trained using different algorithms on different datasets, it becomes important to establish a benchmark to compare the performance of such models on a variety of clinically relevant tasks spanning multiple organs and diseases. In this work, we present a collection of pathology datasets comprising clinical slides associated with clinically relevant endpoints including cancer diagnoses and a variety of biomarkers generated during standard hospital operation from three medical centers. We leverage these datasets to systematically assess the performance of public pathology foundation models and provide insights into best practices for training foundation models and selecting appropriate pretrained models. To enable the community to evaluate their models on our clinical datasets, we make available an automated benchmarking pipeline for external use.

Genetic association studies have identified 44 common genome-wide significant risk loci for late-onset Alzheimer’s disease (LOAD). However, LOAD genetic architecture and prediction are unclear. Here we estimate the optimalP-threshold (P_optimal) of a genetic risk score (GRS) for prediction of LOAD in three independent datasets comprising 676 cases and 35,675 family history proxy cases. We show that the discriminative ability of GRS in LOAD prediction is maximised when selecting a small number of SNPs. Both simulation results and direct estimation indicate that the number of causal common SNPs for LOAD may be less than 100, suggesting LOAD is more oligogenic than polygenic. The best GRS explains approximately 75% of SNP-heritability, and individuals in the top decile of GRS have ten-fold increased odds when compared to those in the bottom decile. In addition, 14 variants are identified that contribute to both LOAD risk and age at onset of LOAD.

Background: Distinct prevalence of inherited genetic predisposition may partially explain the difference of cancer risks across ancestries. Ancestry-specific analyses of germline genomes are required to inform cancer genetic risk and prognosis of diverse populations. Methods: We conducted analyses using germline and somatic sequencing data generated by The Cancer Genome Atlas. Collapsing pathogenic and likely pathogenic variants to cancer predisposition genes (CPG), we analyzed the association between CPGs and cancer types within ancestral groups. We also identified the predisposition-associated two-hit events and gene expression effects in tumors. Results: Genetic ancestry analysis classified the cohort of 9899 cancer cases into individuals of primarily European (N = 8184, 82.7%), African (N = 966, 9.8%), East Asian (N = 649, 6.6%), South Asian (N = 48, 0.5%), Native/Latin American (N = 41, 0.4%), and admixed (N = 11, 0.1%) ancestries. In the African ancestry, we discovered a potentially novel association of BRCA2 in lung squamous cell carcinoma (OR = 41.4 [95% CI, 6.1–275.6]; FDR = 0.002) previously identified in Europeans, along with a known association of BRCA2 in ovarian serous cystadenocarcinoma (OR = 8.5 [95% CI, 1.5–47.4]; FDR = 0.045). In the East Asian ancestry, we discovered one previously known association of BRIP1 in stomach adenocarcinoma (OR = 12.8 [95% CI, 1.8–90.8]; FDR = 0.038). Rare variant burden analysis further identified 7 suggestive associations in African ancestry individuals previously described in European ancestry, including SDHB in pheochromocytoma and paraganglioma, ATM in prostate adenocarcinoma, VHL in kidney renal clear cell carcinoma, FH in kidney renal papillary cell carcinoma, and PTEN in uterine corpus endometrial carcinoma. Most predisposing variants were found exclusively in one ancestry in the TCGA and gnomAD datasets. Loss of heterozygosity was identified for 7 out of the 15 African ancestry carriers of predisposing variants. Further, tumors from the SDHB or BRCA2 carriers showed simultaneous allelic-specific expression and low gene expression of their respective affected genes, and FH splice-site variant carriers showed mis-splicing of FH. Conclusions: While several CPGs are shared across patients, many pathogenic variants are found to be ancestry-specific and trigger somatic effects. Studies using larger cohorts of diverse ancestries are required to pinpoint ancestry-specific genetic predisposition and inform genetic screening strategies.

Young adult cancer has increased in incidence worldwide, but its molecular etiologies remain unclear. We systematically characterize genomic profiles of young adult tumors with ages of onset ≤50 years and compare them to later-onset tumors using over 6,000 cases across 14 cancer types. While young adult tumors generally show lower mutation burdens and comparable copy-number variation rates compared to later-onset cases, they are enriched for multiple driver mutations and copy-number alterations in subtype-specific contexts. Characterization of tumor immune microenvironments reveals pan-cancer patterns of elevated TGF-β response/dendritic cells and lower IFN-γ response/macrophages relative to later-onset tumors, corresponding to age-related responses to immunotherapy in several cancer types. Finally, we identify prevalent clinically actionable events that disproportionally affect young adult or later-onset cases. The resulting catalog of age-related molecular drivers can guide precision diagnostics and treatments for young adult cancer.

Background: Sex has consistently been shown to affect COVID-19 mortality, but it remains unclear how each sex’s clinical outcome may be distinctively shaped by risk factors. Methods: We studied a primary cohort of 4930 patients hospitalized with COVID-19 in a single healthcare system in New York City from the start of the pandemic till August 5, 2020, and a validation cohort of 1645 patients hospitalized with COVID-19 in the same healthcare system from August 5, 2020, to January 13, 2021. Results: Here we show that male sex was independently associated with in-hospital mortality, intubation, and ICU care after adjusting for demographics and comorbidities. Using interaction analysis and sex-stratified models, we found that hypoxia interacted with sex to preferentially increase women’s mortality risk while obesity interacted with sex to preferentially increase women’s risk of intubation and intensive care in our primary cohort. In the validation cohort, we observed that male sex remained an independent risk factor for mortality, but sex-specific interactions were not replicated. Conclusions: We conducted a comprehensive sex-stratified analysis of a large cohort of hospitalized COVID-19 patients, highlighting clinical factors that may contribute to sex differences in the outcome of COVID-19.