Discovery of cell type classification marker genes from single cell RNA sequencing data using NS-Forest.

The use of single-cell transcriptomic technologies that quantitively describe cell transcriptional phenotypes using single cell/nucleus RNA sequencing (scRNA-seq) is revolutionizing our understanding of cell biology, leading to new insights in cell type identification, disease mechanisms, and drug development. The tremendous growth in scRNA-seq data has posed new challenges in efficiently characterizing data-driven cell types and identifying quantifiable marker genes for cell type classification. The use of machine learning and explainable artificial intelligence has emerged as an effective approach to study large-scale scRNA-seq data. NS-Forest is a random forest machine learning-based algorithm that aims to provide a scalable data-driven solution to identify minimum combinations of necessary and sufficient marker genes that capture cell type identity with maximum classification accuracy. Here, we describe the latest version, NS-Forest version 4.0 and its companion Python package (https://github.com/JCVenterInstitute/NSForest), with several enhancements, to select marker gene combinations that exhibit selective expression patterns among closely related cell types and more efficiently perform marker gene selection for large-scale scRNA-seq data atlases with millions of cells. By modularizing the final decision tree step, NS-Forest v4.0 can be used to compare the performance of user-defined marker genes with the NS-Forest computationally-derived marker genes based on the decision tree classifiers. To quantify how well the identified markers exhibit the desired pattern of being exclusively expressed at high levels within their target cell types, we introduce the On-Target Fraction metric that ranges from 0 to1, with a metric of 1 given to markers that are only expressed within their target cell types and not in cells of any other cell types. We have applied NS-Forest v4.0 on scRNA-seq datasets from three human organs, including the brain, kidney, and lung. We observe that NS-Forest v4.0 outperforms previous versions on its ability to identify markers with higher On-Target Fraction values for closely related cell types and outperforms other marker gene selection approaches on the classification performance with significantly higher F-beta scores.

Heterozygosity alters Msh5 binding to meiotic chromosomes in the baker's yeast.

Meiotic crossovers are initiated from programmed DNA double-strand breaks. The Msh4-Msh5 heterodimer is an evolutionarily conserved mismatch repair-related protein complex that promotes meiotic crossovers by stabilizing strand invasion intermediates and joint molecule structures such as Holliday junctions. In vivo studies using homozygous strains of the baker's yeast Saccharomyces cerevisiae (SK1) show that the Msh4-Msh5 complex associates with double-strand break hotspots, chromosome axes, and centromeres. Many organisms have heterozygous genomes that can affect the stability of strand invasion intermediates through heteroduplex rejection of mismatch-containing sequences. To examine Msh4-Msh5 function in a heterozygous context, we performed chromatin immunoprecipitation and sequencing (ChIP-seq) analysis in a rapidly sporulating hybrid S. cerevisiae strain (S288c-sp/YJM789, containing sporulation-enhancing QTLs from SK1), using SNP information to distinguish reads from homologous chromosomes. Overall, Msh5 localization in this hybrid strain was similar to that determined in the homozygous strain (SK1). However, relative Msh5 levels were reduced in regions of high heterozygosity, suggesting that high mismatch densities reduce levels of recombination intermediates to which Msh4-Msh5 binds. Msh5 peaks were also wider in the hybrid background compared to the homozygous strain (SK1). We determined regions containing heteroduplex DNA by detecting chimeric sequence reads with SNPs from both parents. Msh5-bound double-strand break hotspots overlap with regions that have chimeric DNA, consistent with Msh5 binding to heteroduplex-containing recombination intermediates.

Comprehensive mapping of cell fates in microsatellite unstable cancer cells supports dual targeting of WRN and ATR.

Addiction to the WRN helicase is a unique vulnerability of human cancers with high levels of microsatellite instability (MSI-H). However, while prolonged loss of WRN ultimately leads to cell death, little is known about how MSI-H cancers initially respond to acute loss of WRN-knowledge that would be helpful for informing clinical development of WRN targeting therapy, predicting possible resistance mechanisms, and identifying useful biomarkers of successful WRN inhibition. Here, we report the construction of an inducible ligand-mediated degradation system in which the stability of endogenous WRN protein can be rapidly and specifically tuned, enabling us to track the complete sequence of cellular events elicited by acute loss of WRN function. We found that WRN degradation leads to immediate accrual of DNA damage in a replication-dependent manner that curiously did not robustly engage checkpoint mechanisms to halt DNA synthesis. As a result, WRN-degraded MSI-H cancer cells accumulate DNA damage across multiple replicative cycles and undergo successive rounds of increasingly aberrant mitoses, ultimately triggering cell death. Of potential therapeutic importance, we found no evidence of any generalized mechanism by which MSI-H cancers could adapt to near-complete loss of WRN. However, under conditions of partial WRN degradation, addition of low-dose ATR inhibitor significantly increased their combined efficacy to levels approaching full inactivation of WRN. Overall, our results provide the first comprehensive view of molecular events linking upstream inhibition of WRN to subsequent cell death and suggest that dual targeting of WRN and ATR might be a useful strategy for treating MSI-H cancers.

Comprehensive mapping of cell fates in microsatellite unstable cancer cells support dual targe6ng of WRN and ATR.

Addiction to the WRN helicase is a unique vulnerability of human cancers with high levels of microsatellite instability (MSI-H). However, while prolonged loss of WRN ultimately leads to cell death, little is known about how MSI-H cancers initially respond to acute loss of WRN, knowledge that would be helpful for informing clinical development of WRN-targeting therapy, predicting possible resistance mechanisms, and identifying useful biomarkers of successful WRN inhibition. Here, we report the construction of an inducible ligand-mediated degradation system wherein the stability of endogenous WRN protein can be rapidly and specifically tuned, enabling us to track the complete sequence of cellular events elicited by acute loss of WRN function. We find that WRN degradation leads to immediate accrual of DNA damage in a replication-dependent manner that curiously did not robustly engage checkpoint mechanisms to halt DNA synthesis. As a result, WRN-degraded MSI-H cancer cells accumulate DNA damage across multiple replicative cycles and undergo successive rounds of increasingly aberrant mitoses, ultimately triggering cell death. Of potential therapeutic importance, we find no evidence of any generalized mechanism by which MSI-H cancers could adapt to near-complete loss of WRN. However, under conditions of partial WRN degradation, addition of low dose ATR inhibitor significantly increased their combined efficacy to levels approaching full inactivation of WRN. Overall, our results provided the first comprehensive view of molecular events linking upstream inhibition of WRN to subsequent cell death and suggested a potential therapeutical rationale for dual targeting of WRN and ATR.

Regulation of Msh4-Msh5 association with meiotic chromosomes in budding yeast.

In the baker's yeast Saccharomyces cerevisiae, most of the meiotic crossovers are generated through a pathway involving the highly conserved mismatch repair related Msh4-Msh5 complex. To understand the role of Msh4-Msh5 in meiotic crossing over, we determined its genome wide in vivo binding sites in meiotic cells. We show that Msh5 specifically associates with DSB hotspots, chromosome axes, and centromeres on chromosomes. A basal level of Msh5 association with these chromosomal features is observed even in the absence of DSB formation (spo11Δ mutant) at the early stages of meiosis. But efficient binding to DSB hotspots and chromosome axes requires DSB formation and resection and is enhanced by double Holliday junction structures. Msh5 binding is also correlated to DSB frequency and enhanced on small chromosomes with higher DSB and crossover density. The axis protein Red1 is required for Msh5 association with the chromosome axes and DSB hotspots but not centromeres. Although binding sites of Msh5 and other pro-crossover factors like Zip3 show extensive overlap, Msh5 associates with centromeres independent of Zip3. These results on Msh5 localization in wild type and meiotic mutants have implications for how Msh4-Msh5 works with other pro-crossover factors to ensure crossover formation.

Loss of Heterozygosity and Base Mutation Rates Vary Among Saccharomyces cerevisiae Hybrid Strains.

A growing body of evidence suggests that mutation rates exhibit intra-species specific variation. We estimated genome-wide loss of heterozygosity (LOH), gross chromosomal changes, and single nucleotide mutation rates to determine intra-species specific differences in hybrid and homozygous strains of Saccharomyces cerevisiae The mutation accumulation lines of the S. cerevisiae hybrid backgrounds - S288c/YJM789 (S/Y) and S288c/RM11-1a (S/R) were analyzed along with the homozygous diploids RM11, S288c, and YJM145. LOH was extensive in both S/Y and S/R hybrid backgrounds. The S/Y background also showed longer LOH tracts, gross chromosomal changes, and aneuploidy. Short copy number aberrations were observed in the S/R background. LOH data from the S/Y and S/R hybrids were used to construct a LOH map for S288c to identify hotspots. Further, we observe up to a sixfold difference in single nucleotide mutation rates among the S. cerevisiae S/Y and S/R genetic backgrounds. Our results demonstrate LOH is common during mitotic divisions in S. cerevisiae hybrids and also highlight genome-wide differences in LOH patterns and rates of single nucleotide mutations between commonly used S. cerevisiae hybrid genetic backgrounds.

The Neurospora crassa Standard Oak Ridge Background Exhibits Atypically Efficient Meiotic Silencing by Unpaired DNA.

Meiotic silencing by unpaired DNA (MSUD), an RNAi-mediated gene silencing process, is efficient in crosses made in the Neurospora crassa standard Oak Ridge (OR) genetic background. However, MSUD was decidedly less efficient when the OR-derived MSUD testers were crossed with many wild-isolated strains (W), suggesting that either sequence heterozygosity in tester x W crosses suppresses MSUD, or that OR represents the MSUD-conducive extreme in the range of genetic variation in MSUD efficiency. Our results support the latter model. MSUD was less efficient in near-isogenic crosses made in the novel N. crassa B/S1 genetic background, and in N. tetrasperma strain 85. Possibly, in B/S1 and 85, additional regulatory cues, absent from OR, calibrate the MSUD response. A locus in distal chromosome 1R appears to underlie the OR vs. B/S1 difference. Repeat-induced point mutation (RIP) destroys duplicated genes by G:C to A:T mutation of duplicated DNA sequences. Chromosome segment duplications (Dps) dominantly suppress RIP, possibly by titrating out the RIP machinery. In Dp x N crosses, the Dp-borne genes cannot pair properly, hence efficient MSUD, as in OR, silences them and renders the crosses barren. We speculate that the increased productivity engendered by inefficient MSUD enables small duplications to escape RIP.

Genome wide analysis of meiotic recombination in yeast: For a few SNPs more.

Diploid organisms undergo meiosis to produce haploid germ cells. Crossover events during meiosis promote genetic diversity and facilitate accurate chromosome segregation. The baker's yeast Saccharomyces cerevisiae is extensively used as a model for analysis of meiotic recombination. Conventional methods for measuring recombination events in S. cerevisiae have been limited by the number and density of genetic markers. Next generation sequencing (NGS)-based analysis of hybrid yeast genomes bearing thousands of heterozygous single nucleotide polymorphism (SNP) markers has revolutionized analysis of meiotic recombination. By facilitating analysis of marker segregation in the whole genome with unprecedented resolution, this method has resulted in the generation of high-resolution recombination maps in wild-type and meiotic mutants. These studies have provided novel insights into the mechanism of meiotic recombination. In this review, we discuss the methodology, challenges, insights and future prospects of using NGS-based methods for whole genome analysis of meiotic recombination. The objective is to facilitate the use of these high through-put sequencing methods for the analysis of meiotic recombination given their power to provide significant new insights into the process. © 2018 The Authors. IUBMB Life published by Wiley Periodicals, Inc. on behalf of International Union of Biochemistry and Molecular Biology, 70(8):743-752, 2018.

Genome Dynamics of Hybrid Saccharomyces cerevisiae During Vegetative and Meiotic Divisions.

Mutation and recombination are the major sources of genetic diversity in all organisms. In the baker's yeast, all mutation rate estimates are in homozygous background. We determined the extent of genetic change through mutation and loss of heterozygosity (LOH) in a heterozygous Saccharomyces cerevisiae genome during successive vegetative and meiotic divisions. We measured genome-wide LOH and base mutation rates during vegetative and meiotic divisions in a hybrid (S288c/YJM789) S. cerevisiae strain. The S288c/YJM789 hybrid showed nearly complete reduction in heterozygosity within 31 generations of meioses and improved spore viability. LOH in the meiotic lines was driven primarily by the mating of spores within the tetrad. The S288c/YJM789 hybrid lines propagated vegetatively for the same duration as the meiotic lines, showed variable LOH (from 2 to 3% and up to 35%). Two of the vegetative lines with extensive LOH showed frequent and large internal LOH tracts that suggest a high frequency of recombination repair. These results suggest significant LOH can occur in the S288c/YJM789 hybrid during vegetative propagation presumably due to return to growth events. The average base substitution rates for the vegetative lines (1.82 × 10-10 per base per division) and the meiotic lines (1.22 × 10-10 per base per division) are the first genome-wide mutation rate estimates for a hybrid yeast. This study therefore provides a novel context for the analysis of mutation rates (especially in the context of detecting LOH during vegetative divisions), compared to previous mutation accumulation studies in yeast that used homozygous backgrounds.

Modulating Crossover Frequency and Interference for Obligate Crossovers in Saccharomyces cerevisiae Meiosis.

Meiotic crossover frequencies show wide variation among organisms. But most organisms maintain at least one crossover per homolog pair (obligate crossover). In Saccharomyces cerevisiae, previous studies have shown crossover frequencies are reduced in the mismatch repair related mutant mlh3Δ and enhanced in a meiotic checkpoint mutant pch2Δ by up to twofold at specific chromosomal loci, but both mutants maintain high spore viability. We analyzed meiotic recombination events genome-wide in mlh3Δ, pch2Δ, and mlh3Δ pch2Δ mutants to test the effect of variation in crossover frequency on obligate crossovers. mlh3Δ showed ∼30% genome-wide reduction in crossovers (64 crossovers per meiosis) and loss of the obligate crossover, but nonexchange chromosomes were efficiently segregated. pch2Δ showed ∼50% genome-wide increase in crossover frequency (137 crossovers per meiosis), elevated noncrossovers as well as loss of chromosome size dependent double-strand break formation. Meiotic defects associated with pch2∆ did not cause significant increase in nonexchange chromosome frequency. Crossovers were restored to wild-type frequency in the double mutant mlh3Δ pch2Δ (100 crossovers per meiosis), but obligate crossovers were compromised. Genetic interference was reduced in mlh3Δ, pch2Δ, and mlh3Δ pch2Δ. Triple mutant analysis of mlh3Δ pch2Δ with other resolvase mutants showed that most of the crossovers in mlh3Δ pch2Δ are made through the Mus81-Mms4 pathway. These results are consistent with a requirement for increased crossover frequencies in the absence of genetic interference for obligate crossovers. In conclusion, these data suggest crossover frequencies and the strength of genetic interference in an organism are mutually optimized to ensure obligate crossovers.