Search

Objectives: Over the past few years, challenges from the pandemic have led to an explosion of data sharing and algorithmic development efforts in the areas of molecular measurements, clinical data, and digital health. We aim to characterize and describe recent advanced computational approaches in translational bioinformatics across these domains in the context of issues or progress related to equity and inclusion. Methods: We conducted a literature assessment of the trends and approaches in translational bioinformatics in the past few years. Results: We present a review of recent computational approaches across molecular, clinical, and digital realms. We discuss applications of phenotyping, disease subtype characterization, predictive modeling, biomarker discovery, and treatment selection. We consider these methods and applications through the lens of equity and inclusion in biomedicine. Conclusion: Equity and inclusion should be incorporated at every step of translational bioinformatics projects, including project design, data collection, model creation, and clinical implementation. These considerations, coupled with the exciting breakthroughs in big data and machine learning, are pivotal to reach the goals of precision medicine for all.

Recurrent pregnancy loss (RPL), defined as 2 or more pregnancy losses, affects 5-6% of ever-pregnant individuals. Approximately half of these cases have no identifiable explanation. To generate hypotheses about RPL etiologies, we implemented a case-control study comparing the history of over 1,600 diagnoses between RPL and live-birth patients, leveraging the University of California San Francisco (UCSF) and Stanford University electronic health record databases. In total, our study included 8,496 RPL (UCSF: 3,840, Stanford: 4,656) and 53,278 Control (UCSF: 17,259, Stanford: 36,019) patients. Menstrual abnormalities and infertility-associated diagnoses were significantly positively associated with RPL in both medical centers. Age-stratified analysis revealed that the majority of RPL-associated diagnoses had higher odds ratios for patients

Early identification of Alzheimer’s Disease (AD) risk can aid in interventions before disease progression. We demonstrate that electronic health records (EHRs) combined with heterogeneous knowledge networks (e.g., SPOKE) allow for (1) prediction of AD onset and (2) generation of biological hypotheses linking phenotypes with AD. We trained random forest models that predict AD onset with mean AUROC of 0.72 (-7 years) to .81 (-1 day). Top identified conditions from matched cohort trained models include phenotypes with importance across time, early in time, or closer to AD onset. SPOKE networks highlight shared genes between top predictors and AD (e.g., APOE, IL6, TNF, and INS). Survival analysis of top predictors (hyperlipidemia and osteoporosis) in external EHRs validates an increased risk of AD. Genetic colocalization confirms hyperlipidemia and AD association at the APOE locus, and AD with osteoporosis colocalize at a locus close to MS4A6A with a stronger female association.

Although prematurity is the single largest cause of death in children under 5 years of age, the current definition of prematurity, based on gestational age, lacks the precision needed for guiding care decisions. Here, we propose a longitudinal risk assessment for adverse neonatal outcomes in newborns based on a deep learning model that uses electronic health records (EHRs) to predict a wide range of outcomes over a period starting shortly before conception and ending months after birth. By linking the EHRs of the Lucile Packard Children’s Hospital and the Stanford Healthcare Adult Hospital, we developed a cohort of 22,104 mother-newborn dyads delivered between 2014 and 2018. Maternal and newborn EHRs were extracted and used to train a multi-input multitask deep learning model, featuring a long short-term memory neural network, to predict 24 different neonatal outcomes. An additional cohort of 10,250 mother-newborn dyads delivered at the same Stanford Hospitals from 2019 to September 2020 was used to validate the model. Areas under the receiver operating characteristic curve at delivery exceeded 0.9 for 10 of the 24 neonatal outcomes considered and were between 0.8 and 0.9 for 7 additional outcomes. Moreover, comprehensive association analysis identified multiple known associations between various maternal and neonatal features and specific neonatal outcomes. This study used linked EHRs from more than 30,000 mother-newborn dyads and would serve as a resource for the investigation and prediction of neonatal outcomes. An interactive website is available for independent investigators to leverage this unique dataset: https://maternal-child-health-associations.shinyapps.io/shiny_app/ .

Background: Aggressive breast cancers have increased proliferation or metastatic potential and portend a poor prognosis. The ability to identify women at elevated risk of aggressive cancers could have major implications for screening and prevention, yet there are no available tools for predicting aggressive cancer risk. We sought to construct a polygenic risk score (PRS) for aggressive breast cancers by leveraging the associations of single nucleotide polymorphisms (SNPs) with tumor gene expression. We used as our measure of aggressiveness the risk of recurrence score weighted on proliferation (ROR-P), a validated tumor prognostic signature. We hypothesized that known breast cancer susceptibility SNPs would have differential associations with ROR-P, which could then be used to construct a PRS for ROR-P. Methods: We developed our PRS in a case-only analysis of 3 studies containing SNP genotypes and tumor gene expression: The Cancer Genome Atlas, METABRIC, and the I-SPY 2 TRIAL (total n=2,363). We used linear regression models to evaluate individual SNP associations with ROR-P, adjusted for genetic ancestry and study. We then constructed PRS using varying p-value thresholds and used cross-validation to identify the PRS with highest model r2. To assess whether the ROR-P PRS was associated with poor prognosis, we performed survival analysis in two longitudinal cohorts of breast cancer patients: the UK Biobank (women with incident invasive cancers only) and the Pathways Study. These studies included 10,196 total cases with 785 deaths. We built Cox proportional hazards models to evaluate the association between the ROR-P PRS (adjusted for genetic ancestry) and breast cancer-specific survival (BCSS) in both studies. We then performed meta-analysis of the Cox model results. We also constructed joint models containing the ROR-P PRS and a PRS representing the case-case risk of ER-negative vs. ER-positive cancer, PRSER-/ER+. Results: We tested the associations between 226 breast cancer susceptibility SNPs and ROR-P. The best-performing PRS contained 76 SNPs and had a cross-validated r2 of 0.051. In the UK Biobank and Pathways Study, higher ROR-P PRS was associated with worse BCSS, with nearly identical effects observed in each study, HR per standard deviation of 1.13 (95% CI 1.05-1.21, p=9.0x10-4) in meta-analysis. The ROR-P PRS’s effect was minimally attenuated when adjusted for PRSER-/ER+, suggesting that the ROR-P PRS was providing additional prognostic information beyond ER status. Conclusions: We used breast cancer susceptibility SNPs to construct a PRS for ROR-P, a prognostic signature recapitulating aggressiveness, and found the ROR-P PRS to be associated with worse BCSS. Our findings represent an improvement on current PRS for overall breast cancer risk, which preferentially predict cancers with favorable prognosis. Given that aggressive cancers are more likely to present as advanced cancers even among women undergoing routine screening, our findings could potentially identify women who may benefit from more intensive screening. Citation Format: Yiwey Shieh, Jacquelyn Roger, Christina Yau, Denise Wolf, Gillian Hirst, Lamorna Swigart, Scott Huntsman, Donglei Hu, Jovia Nierenberg, Pooja Middha, Rachel Heise, Linda Kachuri, Qianqian Zhu, Song Yao, Christine Ambrosone, Marilyn Kwan, Bette Caan, John Witte, Lawrence Kushi, Laura van ’T. Veer, Laura Esserman, Elad Ziv. Development and testing of a polygenic risk score for breast cancer. Aggressiveness. [abstract]. In: Proceedings of the AACR Special Conference: Precision Prevention, Early Detection, and Interception of Cancer; 2022 Nov 17-19; Austin, TX. Philadelphia (PA): AACR; Can Prev Res 2023;16(1 Suppl): Abstract nr PR008.

Background: Identification of women at elevated risk for highly proliferative, poor prognosis breast cancers could have important implications for screening and prevention. Genome-wide association studies (GWAS) have found >200 single nucleotide polymorphisms (SNPs) associated with breast cancer risk, with many SNPs differentially associated with ER status or intrinsic subtype. We hypothesized that some of these SNPs are preferentially associated with more proliferative tumors, while others are preferentially associated with less proliferative tumors. In this study, we aimed to build a polygenic risk score (PRS) predictive of proliferative tumors, using the GWAS-identified SNPs. Methods: We used data from 3 studies that included array-based SNP genotyping and tumor transcriptomic data: The Cancer Genome Atlas, METABRIC, and the I-SPY 2 Trial (total n=2,467). Our outcome was the risk of recurrence score weighted on proliferation (ROR-P), a validated tumor prognostic signature. Using the breast cancer risk SNPs, we built respective linear regression models to predict ROR-P, with genetic ancestry, study, and ER status as covariates. We performed 5-fold cross-validation and used the model r2 to identify the optimal p-value threshold for including SNPs in the PRS. To decrease uncertainty of our estimates, we performed 100 repeats across the pooled datasets. To test whether this model predicted poor prognosis breast cancers, we first used it to impute ROR-P among genotyped breast cancer cases in UK Biobank (UKB). We then examined the association between “genetically predicted” ROR-P and breast cancer-specific survival using Cox proportional hazards models adjusted for genetic ancestry and age at diagnosis. Results: Associations between 224 breast cancer SNPs and ROR-P were tested. The best-performing model in cross-validation contained 96 SNPs, each associated with ROR-P at p < 0.45, with model r2 0.054. The SNPs with the strongest positive correlations with ROR-P included those discovered in GWAS for HER2-positive and ER-negative cancers, both of which tend to be highly proliferative. Among 7,247 incident cancers in UKB, higher genetically predicted ROR-P was associated with shorter survival, with a per-standard deviation hazard ratio of 1.14 (95% CI 1.05-1.24, p = 0.002). Conclusions: We used breast cancer susceptibility SNPs to construct a PRS fitted to ROR-P, a prognostic signature recapitulating tumor proliferation. This PRS was associated with worse clinical outcomes in breast cancer cases from UKB. Our results suggest that correlations between SNPs and tumor gene expression can be used to “tune” PRS to tumor phenotype, e.g. proliferation. Highly proliferative tumors are more likely to present as advanced cancers even among women getting routine screening. If replicated in other datasets, our findings could be used to identify women who may especially benefit from tailored screening and prevention. Citation Format: Yiwey Shieh, Jacquelyn Roger, Scott Huntsman, Donglei Hu, Jovia L. Nierenberg, Pooja Middha Kapoor, Christina Yau, Gillian Hirst, Laura van 't Veer, Laura Esserman. Development and testing of a polygenic risk score for proliferative breast cancers [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2022; 2022 Apr 8-13. Philadelphia (PA): AACR; Cancer Res 2022;82(12_Suppl):Abstract nr 5883.