Gene regulatory networks (GRNs) govern many core developmental and biological processes underlying human complex traits. Even with broad-scale efforts to characterize the effects of molecular perturbations and interpret gene coexpression, it remains challenging to infer the architecture of gene regulation in a precise and efficient manner. Key properties of GRNs, like hierarchical structure, modular organization, and sparsity, provide both challenges and opportunities for this objective. Here, we seek to better understand properties of GRNs using a new approach to simulate their structure and model their function. We produce realistic network structures with a novel generating algorithm based on insights from small-world network theory, and we model gene expression regulation using stochastic differential equations formulated to accommodate modeling molecular perturbations. With these tools, we systematically describe the effects of gene knockouts within and across GRNs, finding a subset of networks that recapitulate features of a recent genome-scale perturbation study. With deeper analysis of these exemplar networks, we consider future avenues to map the architecture of gene expression regulation using data from cells in perturbed and unperturbed states, finding that while perturbation data are critical to discover specific regulatory interactions, data from unperturbed cells may be sufficient to reveal regulatory programs.

A common theme in causal inference is learning causal relationships between observed variables, also known as causal discovery. This is usually a daunting task, given the large number of candidate causal graphs and the combinatorial nature of the search space. Perhaps for this reason, most research has so far focused on relatively small causal graphs, with up to hundreds of nodes. However, recent advances in fields like biology enable generating experimental data sets with thousands of interventions followed by rich profiling of thousands of variables, raising the opportunity and urgent need for large causal graph models. Here, we introduce the notion of factor directed acyclic graphs ($f$-DAGs) as a way to restrict the search space to non-linear low-rank causal interaction models. Combining this novel structural assumption with recent advances that bridge the gap between causal discovery and continuous optimization, we achieve causal discovery on thousands of variables. Additionally, as a model for the impact of statistical noise on this estimation procedure, we study a model of edge perturbations of the $f$-DAG skeleton based on random graphs and quantify the effect of such perturbations on the $f$-DAG rank. This theoretical analysis suggests that the set of candidate $f$-DAGs is much smaller than the whole DAG space and thus may be more suitable as a search space in the high-dimensional regime where the underlying skeleton is hard to assess. We propose Differentiable Causal Discovery of Factor Graphs (DCD-FG), a scalable implementation of $f$-DAG constrained causal discovery for high-dimensional interventional data. DCD-FG uses a Gaussian non-linear low-rank structural equation model and shows significant improvements compared to state-of-the-art methods in both simulations as well as a recent large-scale single-cell RNA sequencing data set with hundreds of genetic interventions.

Quantification of cell-free DNA (cfDNA) in circulating blood derived from a transplanted organ is a powerful approach to monitoring post-transplant injury. Genome transplant dynamics (GTD) quantifies donor-derived cfDNA (dd-cfDNA) by taking advantage of single-nucleotide polymorphisms (SNPs) distributed across the genome to discriminate donor and recipient DNA molecules. In its current implementation, GTD requires genotyping of both the transplant recipient and donor. However, in practice, donor genotype information is often unavailable. Here, we address this issue by developing an algorithm that estimates dd-cfDNA levels in the absence of a donor genotype. Our algorithm predicts heart and lung allograft rejection with an accuracy that is similar to conventional GTD. We furthermore refined the algorithm to handle closely related recipients and donors, a scenario that is common in bone marrow and kidney transplantation. We show that it is possible to estimate dd-cfDNA in bone marrow transplant patients that are unrelated or that are siblings of the donors, using a hidden Markov model (HMM) of identity-by-descent (IBD) states along the genome. Last, we demonstrate that comparing dd-cfDNA to the proportion of donor DNA in white blood cells can differentiate between relapse and the onset of graft-versus-host disease (GVHD). These methods alleviate some of the barriers to the implementation of GTD, which will further widen its clinical application.

Abstract Motivation: Sequencing-based assays such as ChIP-seq, DNase-seq and MNase-seq have become important tools for genome annotation. In these assays, short sequence reads enriched for loci of interest are mapped to a reference genome to determine their origin. Here, we consider whether false positive peak calls can be caused by particular type of error in the reference genome: multicopy sequences which have been incorrectly assembled and collapsed into a single copy. Results: Using sequencing data from the 1000 Genomes Project, we systematically scanned the human genome for regions of high sequencing depth. These regions are highly enriched for erroneously inferred transcription factor binding sites, positions of nucleosomes and regions of open chromatin. We suggest a simple masking procedure to remove these regions and reduce false positive calls. Availability: Files for masking out these regions are available at eqtl.uchicago.edu Contact: [email protected]; [email protected]; [email protected]; [email protected] Supplementary information: Supplementary data are available at Bioinformatics online.

BACKGROUND: Counting k-mers (substrings of length k in DNA sequence data) is an essential component of many methods in bioinformatics, including for genome and transcriptome assembly, for metagenomic sequencing, and for error correction of sequence reads. Although simple in principle, counting k-mers in large modern sequence data sets can easily overwhelm the memory capacity of standard computers. In current data sets, a large fraction-often more than 50%-of the storage capacity may be spent on storing k-mers that contain sequencing errors and which are typically observed only a single time in the data. These singleton k-mers are uninformative for many algorithms without some kind of error correction. RESULTS: We present a new method that identifies all the k-mers that occur more than once in a DNA sequence data set. Our method does this using a Bloom filter, a probabilistic data structure that stores all the observed k-mers implicitly in memory with greatly reduced memory requirements. We then make a second sweep through the data to provide exact counts of all nonunique k-mers. For example data sets, we report up to 50% savings in memory usage compared to current software, with modest costs in computational speed. This approach may reduce memory requirements for any algorithm that starts by counting k-mers in sequence data with errors. CONCLUSIONS: A reference implementation for this methodology, BFCounter, is written in C++ and is GPL licensed. It is available for free download at http://pritch.bsd.uchicago.edu/bfcounter.html.

MOTIVATION: Next-generation sequencing has become an important tool for genome-wide quantification of DNA and RNA. However, a major technical hurdle lies in the need to map short sequence reads back to their correct locations in a reference genome. Here, we investigate the impact of SNP variation on the reliability of read-mapping in the context of detecting allele-specific expression (ASE). RESULTS: We generated 16 million 35 bp reads from mRNA of each of two HapMap Yoruba individuals. When we mapped these reads to the human genome we found that, at heterozygous SNPs, there was a significant bias toward higher mapping rates of the allele in the reference sequence, compared with the alternative allele. Masking known SNP positions in the genome sequence eliminated the reference bias but, surprisingly, did not lead to more reliable results overall. We find that even after masking, approximately 5-10% of SNPs still have an inherent bias toward more effective mapping of one allele. Filtering out inherently biased SNPs removes 40% of the top signals of ASE. The remaining SNPs showing ASE are enriched in genes previously known to harbor cis-regulatory variation or known to show uniparental imprinting. Our results have implications for a variety of applications involving detection of alternate alleles from short-read sequence data. AVAILABILITY: Scripts, written in Perl and R, for simulating short reads, masking SNP variation in a reference genome and analyzing the simulation output are available upon request from JFD. Raw short read data were deposited in GEO (http://www.ncbi.nlm.nih.gov/geo/) under accession number GSE18156. CONTACT: [email protected]; [email protected]; [email protected]; [email protected] SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Summary: The program package TreeLD implements a unified approach to association mapping and fine mapping of complex trait loci and a novel approach to visualizing association data, based on an inferred ancestry of the sample. Fundamentally, the TreeLD approach is based on the idea that the evidence for association at a particular position is contained in the ancestral tree relating the sampled chromosomes at that position. TreeLD provides an easy-to-use interface and can be applied to case–control, TDT trio and quantitative trait data. Availability: The program TreeLD is available on the Internet from http://pritch.bsd.uchicago.edu/software.html Contact: [email protected]