Research articles

Cluster hierarchy optimization by iterative random forests (CHOIR) offers a robust and accurate method to identify cell clusters across a variety of single-cell resolution data with statistical support.

HDL-L is an extension of the high-definition likelihood method that enables local heritability and genetic correlation analysis with higher accuracy and computational efficiency than LAVA.

This study develops family-based genome-wide association study methods that maximize power in homogeneous samples through inclusion of singletons and in diverse samples by using all available parental genotypes.

Simultaneous profiling of the genome, methylome, epigenome and transcriptome using single-molecule chromatin fiber sequencing and multiplexed arrays isoform sequencing identifies the genetic and molecular basis of an undiagnosed Mendelian disease case with an X;13-balanced translocation.

Quickdraws is a mixed-model association tool with a noninfinitesimal prior for analyzing binary and quantitative traits, using a scalable variational inference that allows analysis of biobank-scale cohorts.

Programmable Enrichment via RNA FlowFISH by sequencing (PERFF-seq) isolates rare cells based on RNA marker transcripts for single-cell RNA sequencing profiling of complex tissues, with applicability to a broad variety of samples and cell types.

Generalized binary covariance decomposition (GBCD) applies empirical Bayes matrix factorization to identify shared and sample-specific gene expression signatures in single-cell RNA sequencing data, and can more accurately capture inter- and intrasample heterogeneity than existing methods.

ChIP-DIP (ChIP done in parallel) is a highly multiplex assay for protein–DNA binding, scalable to hundreds of proteins including modified histones, chromatin regulators and transcription factors, offering a refined view of the cis-regulatory code.

Deep rare variant association testing (DeepRVAT) is a deep set neural network model that flexibly integrates rare variant annotations into a trait-agnostic gene impairment score. These scores improve association testing and polygenic risk prediction.

Microfluidics-assisted grid chips for spatial transcriptome sequencing (MAGIC-seq) is a spatial transcriptomics method combining multiple-grid microfluidic design and prefabricated DNA arrays for increased throughput and reduced cost, with applications for large fields of view and 3D spatial mapping.

This study presents a synthetic surrogate (SynSurr) method for imputing missing phenotypes in biobank datasets. Joint analysis of the partially observed and imputed surrogate phenotype improves power in genome-wide association studies while being robust to imputation errors.

GSA-MiXeR models gene heritability and variant linkage disequilibrium for improved gene set enrichment testing. GSA-MiXeR implicates relevant sets of fewer than ten genes in schizophrenia, providing more nuanced insights into trait biology.

Two low-input tagmentation-based long-read sequencing methods, single-molecule real-time sequencing by tagmentation (SMRT-Tag), which identifies genetic variation and CpG methylation, and single-molecule adenine-methylated oligonucleosome sequencing assay by tagmentation (SAMOSA-Tag), which detects chromatin accessibility, are presented. Application of SAMOSA-Tag to prostate cancer patient-derived xenograft samples identifies metastasis-associated epigenomic alterations.

MuSiCal is a mutational signature analysis tool combining minimum-volume nonnegative matrix factorization with other algorithmic innovations. Applied to PCAWG data, MuSiCal gives more accurate results, including resolving ambiguous flat signatures.

Causal-TWAS (cTWAS) is a statistical framework that adjusts for genetic confounders in transcriptome-wide association studies. Application of cTWAS on common traits leads to reliable detection of candidate causal genes.

GIFT fine-maps candidate causal genes in a transcription-wide association study by conditioning on predicted expression of nearby genes, leading to improved statistical power and enhanced mapping resolution when applied to complex traits.

A new method allows selection of matched controls from an external pool of samples without genotype sharing. This method has been implemented in an online repository containing 39,472 exome sequencing controls that can be used for association analyses.

MESuSiE extends fine-mapping approaches to multi-ancestry analysis using LD-aware bivariate normal mixture models with a variational algorithm to identify shared and ancestry-specific causal variants.

A powerful Bayesian method, BridgePRS, leverages shared genetic effects across ancestries to increase polygenic risk score portability in non-European populations.

AutoComplete is a deep learning-based method that imputes missing phenotypes in population-scale biobank datasets, increasing effective sample sizes and improving power for genetic discoveries in genome-wide association studies.