A deep-learning framework to predict cancer treatment response from histopathology images through imputed transcriptomics.

Advances in artificial intelligence have paved the way for leveraging hematoxylin and eosin-stained tumor slides for precision oncology. We present ENLIGHT-DeepPT, an indirect two-step approach consisting of (1) DeepPT, a deep-learning framework that predicts genome-wide tumor mRNA expression from slides, and (2) ENLIGHT, which predicts response to targeted and immune therapies from the inferred expression values. We show that DeepPT successfully predicts transcriptomics in all 16 The Cancer Genome Atlas cohorts tested and generalizes well to two independent datasets. ENLIGHT-DeepPT successfully predicts true responders in five independent patient cohorts involving four different treatments spanning six cancer types, with an overall odds ratio of 2.28 and a 39.5% increased response rate among predicted responders versus the baseline rate. Notably, its prediction accuracy, obtained without any training on the treatment data, is comparable to that achieved by directly predicting the response from the images, which requires specific training on the treatment evaluation cohorts.

Prediction of DNA methylation-based tumor types from histopathology in central nervous system tumors with deep learning.

Precision in the diagnosis of diverse central nervous system (CNS) tumor types is crucial for optimal treatment. DNA methylation profiles, which capture the methylation status of thousands of individual CpG sites, are state-of-the-art data-driven means to enhance diagnostic accuracy but are also time consuming and not widely available. Here, to address these limitations, we developed Deep lEarning from histoPathoLOgy and methYlation (DEPLOY), a deep learning model that classifies CNS tumors to ten major categories from histopathology. DEPLOY integrates three distinct components: the first classifies CNS tumors directly from slide images ('direct model'), the second initially generates predictions for DNA methylation beta values, which are subsequently used for tumor classification ('indirect model'), and the third classifies tumor types directly from routinely available patient demographics. First, we find that DEPLOY accurately predicts beta values from histopathology images. Second, using a ten-class model trained on an internal dataset of 1,796 patients, we predict the tumor categories in three independent external test datasets including 2,156 patients, achieving an overall accuracy of 95% and balanced accuracy of 91% on samples that are predicted with high confidence. These results showcase the potential future use of DEPLOY to assist pathologists in diagnosing CNS tumors within a clinically relevant short time frame.

Prediction of cancer treatment response from histopathology images through imputed transcriptomics.

Advances in artificial intelligence have paved the way for leveraging hematoxylin and eosin (H&E)-stained tumor slides for precision oncology. We present ENLIGHT-DeepPT, an approach for predicting response to multiple targeted and immunotherapies from H&E-slides. In difference from existing approaches that aim to predict treatment response directly from the slides, ENLIGHT-DeepPT is an indirect two-step approach consisting of (1) DeepPT, a new deep-learning framework that predicts genome-wide tumor mRNA expression from slides, and (2) ENLIGHT, which predicts response based on the DeepPT inferred expression values. DeepPT successfully predicts transcriptomics in all 16 TCGA cohorts tested and generalizes well to two independent datasets. Our key contribution is showing that ENLIGHT-DeepPT successfully predicts true responders in five independent patients' cohorts involving four different treatments spanning six cancer types with an overall odds ratio of 2.44, increasing the baseline response rate by 43.47% among predicted responders, without the need for any treatment data for training. Furthermore, its prediction accuracy on these datasets is comparable to a supervised approach predicting the response directly from the images, which needs to be trained and tested on the same cohort. ENLIGHT-DeepPT future application could provide clinicians with rapid treatment recommendations to an array of different therapies and importantly, may contribute to advancing precision oncology in developing countries.

Inferring Species Compositions of Complex Fungal Communities from Long- and Short-Read Sequence Data.

The kingdom Fungi is highly diverse in morphology and ecosystem function. Yet fungi are challenging to characterize as they can be difficult to culture and morphologically indistinct. Overall, their description and analysis lag far behind other microbes such as bacteria. Classification of species via high-throughput sequencing is increasingly becoming the norm for pathogen detection, microbiome studies, and environmental monitoring. With the rapid development of sequencing technologies, however, standardized procedures for taxonomic assignment of long sequence reads have not yet been well established. Focusing on nanopore sequencing technology, we compared classification and community composition analysis pipelines using shotgun and amplicon sequencing data generated from mock communities comprising 43 fungal species. We show that regardless of the sequencing methodology used, the highest accuracy of species identification was achieved by sequence alignment against a fungal-specific database. During the assessment of classification algorithms, we found that applying cutoffs to the query coverage of each read or contig significantly improved the classification accuracy and community composition analysis without major data loss. We also generated draft genome assemblies for three fungal species from nanopore data which were absent from genome databases. Our study improves sequence-based classification and estimation of relative sequence abundance using real fungal community data and provides a practical guide for the design of metagenomics analyses focusing on fungi. IMPORTANCE Our study is unique in that it provides an in-depth comparative study of a real-life complex fungal community analyzed with multiple long- and short-read sequencing approaches. These technologies and their application are currently of great interest to diverse biologists as they seek to characterize the community compositions of microbiomes. Although great progress has been made on bacterial community compositions, microbial eukaryotes such as fungi clearly lag behind. Our study provides a detailed breakdown of strategies to improve species identification with immediate relevance to real-world studies. We find that real-life data sets do not always behave as expected, distinct from reports based on simulated data sets.

Physical separation of haplotypes in dikaryons allows benchmarking of phasing accuracy in Nanopore and HiFi assemblies with Hi-C data.

BACKGROUND: Most animals and plants have more than one set of chromosomes and package these haplotypes into a single nucleus within each cell. In contrast, many fungal species carry multiple haploid nuclei per cell. Rust fungi are such species with two nuclei (karyons) that contain a full set of haploid chromosomes each. The physical separation of haplotypes in dikaryons means that, unlike in diploids, Hi-C chromatin contacts between haplotypes are false-positive signals. RESULTS: We generate the first chromosome-scale, fully-phased assembly for the dikaryotic leaf rust fungus Puccinia triticina and compare Nanopore MinION and PacBio HiFi sequence-based assemblies. We show that false-positive Hi-C contacts between haplotypes are predominantly caused by phase switches rather than by collapsed regions or Hi-C read mis-mappings. We introduce a method for phasing of dikaryotic genomes into the two haplotypes using Hi-C contact graphs, including a phase switch correction step. In the HiFi assembly, relatively few phase switches occur, and these are predominantly located at haplotig boundaries and can be readily corrected. In contrast, phase switches are widespread throughout the Nanopore assembly. We show that haploid genome read coverage of 30-40 times using HiFi sequencing is required for phasing of the leaf rust genome, with 0.7% heterozygosity, and that HiFi sequencing resolves genomic regions with low heterozygosity that are otherwise collapsed in the Nanopore assembly. CONCLUSIONS: This first Hi-C based phasing pipeline for dikaryons and comparison of long-read sequencing technologies will inform future genome assembly and haplotype phasing projects in other non-haploid organisms.

The stem rust fungus Puccinia graminis f. sp. tritici induces centromeric small RNAs during late infection that are associated with genome-wide DNA methylation.

BACKGROUND: Silencing of transposable elements (TEs) is essential for maintaining genome stability. Plants use small RNAs (sRNAs) to direct DNA methylation to TEs (RNA-directed DNA methylation; RdDM). Similar mechanisms of epigenetic silencing in the fungal kingdom have remained elusive. RESULTS: We use sRNA sequencing and methylation data to gain insight into epigenetics in the dikaryotic fungus Puccinia graminis f. sp. tritici (Pgt), which causes the devastating stem rust disease on wheat. We use Hi-C data to define the Pgt centromeres and show that they are repeat-rich regions (~250 kb) that are highly diverse in sequence between haplotypes and, like in plants, are enriched for young TEs. DNA cytosine methylation is particularly active at centromeres but also associated with genome-wide control of young TE insertions. Strikingly, over 90% of Pgt sRNAs and several RNAi genes are differentially expressed during infection. Pgt induces waves of functionally diversified sRNAs during infection. The early wave sRNAs are predominantly 21 nts with a 5' uracil derived from genes. In contrast, the late wave sRNAs are mainly 22-nt sRNAs with a 5' adenine and are strongly induced from centromeric regions. TEs that overlap with late wave sRNAs are more likely to be methylated, both inside and outside the centromeres, and methylated TEs exhibit a silencing effect on nearby genes. CONCLUSIONS: We conclude that rust fungi use an epigenetic silencing pathway that might have similarity with RdDM in plants. The Pgt RNAi machinery and sRNAs are under tight temporal control throughout infection and might ensure genome stability during sporulation.

Case-comparison study protocol for gauging effects of neighbourhood trends and sickness: examining the perceptions of transit-Induced gentrification in Prince George's County.

INTRODUCTION: Impoverished neighbourhoods and communities of colour often bear the brunt of unintended transit-oriented development (TOD) impacts. These impacts have been known to come in the form of transit-induced gentrification (TIG), a socioeconomic by-product of TOD defined as a phenomenon that occurs when the provision of transit service, particularly light rail transit (LRT), 'up-scales' nearby neighbourhood(s) and displaces existing residents. Consequently, TIG or even the perception of TIG can impact health outcomes (eg, anxiety) and social determinants of health (SDOH) (eg, crime). METHODS/ANALYSIS: In 2022, the purple line (PL), a 16.2 mile LRT line, is opening in Prince George's County, Maryland, a suburb of Washington, DC, comprised of over 80% African American and Hispanic residents. By taking advantage of this natural experiment, we are proposing the GENTS (Gauging Effects of Neighborhood Trends and Sickness: Examining the Perceptions of Transit-Induced Gentrification in Prince George's County) Study in order to evaluate perceived TIG and associated health outcome and SDOH changes, at two points in time, among Prince George's County adults in a prospective case-comparison design during the pre-PL LRT period. Descriptive analysis and latent growth curve modelling will be used to examine these changes over time. ETHICS AND DISSEMINATION: Ethics approval has been granted by the University of Maryland Institutional Review Board. The GENTS Study will identify temporal changes in perceived TIG, health outcomes and SDOH among case and comparison residents before the completion and operation of the PL LRT, an under researched period of TOD. The dissemination of GENTS Study findings will be able to address research questions and policy issues that are specifically tailored to PG County while also providing more effective procedural solutions for other regions undergoing TOD and TIG risks.

The Prevalence and Impact of Model Violations in Phylogenetic Analysis.

In phylogenetic inference, we commonly use models of substitution which assume that sequence evolution is stationary, reversible, and homogeneous (SRH). Although the use of such models is often criticized, the extent of SRH violations and their effects on phylogenetic inference of tree topologies and edge lengths are not well understood. Here, we introduce and apply the maximal matched-pairs tests of homogeneity to assess the scale and impact of SRH model violations on 3,572 partitions from 35 published phylogenetic data sets. We show that roughly one-quarter of all the partitions we analyzed (23.5%) reject the SRH assumptions, and that for 25% of data sets, tree topologies inferred from all partitions differ significantly from topologies inferred using the subset of partitions that do not reject the SRH assumptions. This proportion increases when comparing trees inferred using the subset of partitions that rejects the SRH assumptions, to those inferred from partitions that do not reject the SRH assumptions. These results suggest that the extent and effects of model violation in phylogenetics may be substantial. They highlight the importance of testing for model violations and possibly excluding partitions that violate models prior to tree reconstruction. Our results also suggest that further effort in developing models that do not require SRH assumptions could lead to large improvements in the accuracy of phylogenomic inference. The scripts necessary to perform the analysis are available in https://github.com/roblanf/SRHtests, and the new tests we describe are available as a new option in IQ-TREE (http://www.iqtree.org).

Functional rare and low frequency variants in BLK and BANK1 contribute to human lupus.

Systemic lupus erythematosus (SLE) is the prototypic systemic autoimmune disease. It is thought that many common variant gene loci of weak effect act additively to predispose to common autoimmune diseases, while the contribution of rare variants remains unclear. Here we describe that rare coding variants in lupus-risk genes are present in most SLE patients and healthy controls. We demonstrate the functional consequences of rare and low frequency missense variants in the interacting proteins BLK and BANK1, which are present alone, or in combination, in a substantial proportion of lupus patients. The rare variants found in patients, but not those found exclusively in controls, impair suppression of IRF5 and type-I IFN in human B cell lines and increase pathogenic lymphocytes in lupus-prone mice. Thus, rare gene variants are common in SLE and likely contribute to genetic risk.

Accuracy of Genomic Prediction for Foliar Terpene Traits in Eucalyptus polybractea.

Unlike agricultural crops, most forest species have not had millennia of improvement through phenotypic selection, but can contribute energy and material resources and possibly help alleviate climate change. Yield gains similar to those achieved in agricultural crops over millennia could be made in forestry species with the use of genomic methods in a much shorter time frame. Here we compare various methods of genomic prediction for eight traits related to foliar terpene yield in Eucalyptus polybractea, a tree grown predominantly for the production of Eucalyptus oil. The genomic markers used in this study are derived from shallow whole genome sequencing of a population of 480 trees. We compare the traditional pedigree-based additive best linear unbiased predictors (ABLUP), genomic BLUP (GBLUP), BayesB genomic prediction model, and a form of GBLUP based on weighting markers according to their influence on traits (BLUP|GA). Predictive ability is assessed under varying marker densities of 10,000, 100,000 and 500,000 SNPs. Our results show that BayesB and BLUP|GA perform best across the eight traits. Predictive ability was higher for individual terpene traits, such as foliar α-pinene and 1,8-cineole concentration (0.59 and 0.73, respectively), than aggregate traits such as total foliar oil concentration (0.38). This is likely a function of the trait architecture and markers used. BLUP|GA was the best model for the two biomass related traits, height and 1 year change in height (0.25 and 0.19, respectively). Predictive ability increased with marker density for most traits, but with diminishing returns. The results of this study are a solid foundation for yield improvement of essential oil producing eucalypts. New markets such as biopolymers and terpene-derived biofuels could benefit from rapid yield increases in undomesticated oil-producing species.

Bias-Variance Tradeoffs in Recombination Rate Estimation.

In 2013, we and coauthors published a paper characterizing rates of recombination within the 2.1-megabase garnet-scalloped (g-sd) region of the Drosophila melanogaster X chromosome. To extract the signal of recombination in our high-throughput sequence data, we adopted a nonparametric smoothing procedure, reducing variance at the cost of biasing individual recombination rates. In doing so, we sacrificed accuracy to gain precision-precision that allowed us to detect recombination rate heterogeneity. Negotiating the bias-variance tradeoff enabled us to resolve significant variation in the frequency of crossing over across the garnet-scalloped region.

Sporadic, Global Linkage Disequilibrium Between Unlinked Segregating Sites.

Demographic, genetic, or stochastic factors can lead to perfect linkage disequilibrium (LD) between alleles at two loci without respect to the extent of their physical distance, a phenomenon that Lawrence et al. (2005a) refer to as "genetic indistinguishability." This phenomenon can complicate genotype-phenotype association testing by hindering the ability to localize causal alleles, but has not been thoroughly explored from a theoretical perspective or using large, dense whole-genome polymorphism data sets. We derive a simple theoretical model of the prevalence of genetic indistinguishability between unlinked loci and verify its accuracy via simulation. We show that sample size and minor allele frequency are the major determinants of the prevalence of perfect LD between unlinked loci but that demographic factors, such as deviations from random mating, can produce significant effects as well. Finally, we quantify this phenomenon in three model organisms and find thousands of pairs of moderate-frequency ([Formula: see text]) genetically indistinguishable variants in relatively large data sets. These results clarify a previously underexplored population genetic phenomenon with important implications for association studies and define conditions under which it is likely to manifest.

Genetic basis of transcriptome diversity in Drosophila melanogaster.

Understanding how DNA sequence variation is translated into variation for complex phenotypes has remained elusive but is essential for predicting adaptive evolution, for selecting agriculturally important animals and crops, and for personalized medicine. Gene expression may provide a link between variation in DNA sequence and organismal phenotypes, and its abundance can be measured efficiently and accurately. Here we quantified genome-wide variation in gene expression in the sequenced inbred lines of the Drosophila melanogaster Genetic Reference Panel (DGRP), increasing the annotated Drosophila transcriptome by 11%, including thousands of novel transcribed regions (NTRs). We found that 42% of the Drosophila transcriptome is genetically variable in males and females, including the NTRs, and is organized into modules of genetically correlated transcripts. We found that NTRs often were negatively correlated with the expression of protein-coding genes, which we exploited to annotate NTRs functionally. We identified regulatory variants for the mean and variance of gene expression, which have largely independent genetic control. Expression quantitative trait loci (eQTLs) for the mean, but not for the variance, of gene expression were concentrated near genes. Notably, the variance eQTLs often interacted epistatically with local variants in these genes to regulate gene expression. This comprehensive characterization of population-scale diversity of transcriptomes and its genetic basis in the DGRP is critically important for a systems understanding of quantitative trait variation.

Recent progress and open questions in Drosophila dosage compensation.

Sexual dimorphism is observed in many traits across diverse taxa, and often it is quite extreme. Within a species, individuals of opposing sex can appear strikingly different, reflecting differences at the molecular level that may be similarly striking. Among the most extreme cases of such molecular sexual dimorphism is the quantity of sex chromosomes that each sex possesses. Hemizygous sex chromosomes are common to many species, and various mechanisms have evolved to regulate transcriptional activity to ensure appropriate sex chromosome-to-autosome gene expression stoichiometry. Among the most thoroughly investigated of these mechanisms is Drosophila melanogaster's male-specific lethal (MSL) complex-mediated dosage compensation. In Drosophila, the male X chromosome transcription is upregulated approximately two-fold in somatic tissues to counterbalance the effects of sex chromosome hemizygosity on transcript abundance. Despite dramatic advances in our understanding of the Drosophila dosage compensation, many questions remain unanswered, and our understanding of its molecular underpinnings remains incomplete. In this review, we synthesize recent progress in the field as a means to highlight open questions, including how the MSL complex targets the X chromosome, how dosage compensation has shaped evolution of X-linked genes, and the degree to which MSL complex-mediated dosage compensation varies in activity across somatic tissues.

The 100-genomes strains, an S. cerevisiae resource that illuminates its natural phenotypic and genotypic variation and emergence as an opportunistic pathogen.

Saccharomyces cerevisiae, a well-established model for species as diverse as humans and pathogenic fungi, is more recently a model for population and quantitative genetics. S. cerevisiae is found in multiple environments-one of which is the human body-as an opportunistic pathogen. To aid in the understanding of the S. cerevisiae population and quantitative genetics, as well as its emergence as an opportunistic pathogen, we sequenced, de novo assembled, and extensively manually edited and annotated the genomes of 93 S. cerevisiae strains from multiple geographic and environmental origins, including many clinical origin strains. These 93 S. cerevisiae strains, the genomes of which are near-reference quality, together with seven previously sequenced strains, constitute a novel genetic resource, the "100-genomes" strains. Our sequencing coverage, high-quality assemblies, and annotation provide unprecedented opportunities for detailed interrogation of complex genomic loci, examples of which we demonstrate. We found most phenotypic variation to be quantitative and identified population, genotype, and phenotype associations. Importantly, we identified clinical origin associations. For example, we found that an introgressed PDR5 was present exclusively in clinical origin mosaic group strains; that the mosaic group was significantly enriched for clinical origin strains; and that clinical origin strains were much more copper resistant, suggesting that copper resistance contributes to fitness in the human host. The 100-genomes strains are a novel, multipurpose resource to advance the study of S. cerevisiae population genetics, quantitative genetics, and the emergence of an opportunistic pathogen.

X-to-autosome expression and msl-2 transcript abundance correlate among Drosophila melanogaster somatic tissues.

In Drosophila melanogaster, the male-specific lethal (MSL) complex has been studied extensively for its role in upregulating male X-linked genes. Recent advances in high-throughput technologies have improved our understanding of how the MSL complex mediates dosage compensation through chromosome-wide chromatin modifications. Most studies, however, have focused on cell line models that cannot reflect any potential heterogeneity of in vivo dosage compensation. Comparisons between cell line and organismal gene-level dosage compensation upregulation suggest the possibility of variation in MSL complex activity among somatic tissues. We hypothesize the degree, up to but not exceeding 2-fold, to which the MSL complex upregulates male X-linked genes varies quantitatively by tissue type. In this model, MSL complex abundance acts as a rheostat to control the extent of upregulation. Using publicly available expression data, we provide evidence for our model in Drosophila somatic tissues. Specifically, we find X-to-autosome expression correlates with the tissue-specific expression of msl-2 which encodes an essential male-specific component of the MSL complex. This result suggests MSL complex mediated dosage compensation varies quantitatively by tissue type. Furthermore, this result has consequences for models explaining the organismal-scale molecular and evolutionary consequences of MSL-mediated dosage compensation.

Enhanced diversity and aflatoxigenicity in interspecific hybrids of Aspergillus flavus and Aspergillus parasiticus.

Aspergillus flavus and A. parasiticus are the two most important aflatoxin-producing fungi responsible for the contamination of agricultural commodities worldwide. Both species are heterothallic and undergo sexual reproduction in laboratory crosses. Here we examine the possibility of interspecific matings between A. flavus and A. parasiticus. These species can be distinguished morphologically and genetically, as well as by their mycotoxin profiles. Aspergillus flavus produces both B aflatoxins and cyclopiazonic acid (CPA), B aflatoxins or CPA alone, or neither mycotoxin; Aspergillus parasiticus produces B and G aflatoxins or the aflatoxin precursor O-methylsterigmatocystin, but not CPA. Only four of forty-five attempted interspecific crosses between opposite mating types of A. flavus and A. parasiticus were fertile and produced viable ascospores. Single ascospore strains from each cross were shown to be recombinant hybrids using multilocus genotyping and array comparative genome hybridization. Conidia of parents and their hybrid progeny were haploid and predominantly monokaryons and dikaryons based on flow cytometry. Multilocus phylogenetic inference showed that experimental hybrid progeny were grouped with naturally occurring A. flavus L strain and A. parasiticus. Higher total aflatoxin concentrations in some F1 progeny strains compared to midpoint parent aflatoxin levels indicate synergism in aflatoxin production; moreover, three progeny strains synthesized G aflatoxins that were not produced by the parents, and there was evidence of allopolyploidization in one strain. These results suggest that hybridization is an important diversifying force resulting in the genesis of novel toxin profiles in these agriculturally important fungi.

Structural properties of the minimum cut of partially-supplied graphs.

It is well known that information about the structure of a graph is contained within its minimum cut. Here we investigate how the minimum cut of one graph informs the structure of a second, related graph. We consider pairs of graphs G and H, with respective Laplacian matrices L and M, and call H partially supplied provided M is a Schur complement of L. Our results show how the minimum cut of H relates to the structure of the larger graph G.

No evidence for a global male-specific lethal complex-mediated dosage compensation contribution to the demasculinization of the Drosophila melanogaster X chromosome.

In Drosophila melanogaster males, the expression of X-linked genes is regulated by mechanisms that operate on a chromosomal scale. One such mechanism, male-specific lethal complex-dependent X-linked dosage compensation, is thought to broadly enhance the expression of male X-linked genes through two-fold transcriptional upregulation. The evolutionary consequences of this form of dosage compensation are not well understood, particularly with regard to genes more highly expressed in males. It has been observed the X chromosome arrangement of these male-biased genes is non-random, consistent with what one might expect if there is a selective advantage for male-biased genes to avoid dosage compensation. Separately, it has been noted that the male-specific lethal complex and its dosage compensation mechanism appear absent in some male tissues, thus providing a control for the selection hypothesis. Here we utilized publicly available datasets to reassess the arrangement of X-linked male-biased expressed genes after accounting for expression in tissues not dosage compensated by the male-specific lethal complex. Our results do not corroborate previous observations supporting organismal-wide detrimental effects by dosage compensation on X-linked male-biased expressed genes. We instead find no evidence that dosage compensation has played a role in the arrangement of dosage compensated male-biased genes on the X chromosome.

Genome-wide analysis in Drosophila reveals age-specific effects of SNPs on fitness traits.

Most organisms exhibit senescence; a decline in physiological function with age. In nature, rates of senescence vary extensively among individuals and this variation has a significant genetic component; however, we know little about the genes underlying senescence. Here we show the first evidence that individual alleles influence fecundity in an age-specific manner and so the genetic basis of natural variation in fecundity changes dramatically with age. We complete a genome-wide association to identify single-nucleotide polymorphisms (SNPs) affecting lifespan and age-specific fecundity using the Drosophila melanogaster Genetic Reference Panel. We identify 1,031 SNPs affecting fecundity and 52 influencing lifespan. Only one SNP is associated with both early- and late-age fecundity. The age-specific effect of candidate genes on fecundity is validated using RNA interference. In addition, there is a dramatic increase in the number of SNPs influencing fecundity with age. This result provides support for the mutation accumulation theory of aging.