Ensemble learning for classifying single-cell data and projection across reference atlases.

SUMMARY: Single-cell data are being generated at an accelerating pace. How best to project data across single-cell atlases is an open problem. We developed a boosted learner that overcomes the greatest challenge with status quo classifiers: low sensitivity, especially when dealing with rare cell types. By comparing novel and published data from distinct scRNA-seq modalities that were acquired from the same tissues, we show that this approach preserves cell-type labels when mapping across diverse platforms. AVAILABILITY AND IMPLEMENTATION: https://github.com/diazlab/ELSA. CONTACT: aaron.diaz@ucsf.edu. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Ten simple rules for an inclusive summer coding program for non-computer-science undergraduates.

Since 2015, we have run a free 9-week summer program that provides non-computer science (CS) undergraduates at San Francisco State University (SFSU) with experience in coding and doing research. Undergraduate research experiences remain very limited at SFSU and elsewhere, so the summer program provides opportunities for many more students beyond the mentoring capacity of our university laboratories. In addition, we were concerned that many students from historically underrepresented (HU) groups may be unable to take advantage of traditional summer research programs because these programs require students to relocate or be available full time, which is not feasible for students who have family, work, or housing commitments. Our program, which is local and part-time, serves about 5 times as many students as a typical National Science Foundation (NSF) Research Experiences for Undergraduates (REU) program, on a smaller budget. Based on our experiences, we present 10 simple rules for busy faculty who want to create similar programs to engage non-CS HU undergraduates in computational research. Note that while some of the strategies we implement are based on evidence-based publications in the social sciences or education research literature, the original suggestions we make here are based on our trial-and-error experiences, rather than formal hypothesis testing.

ATRX regulates glial identity and the tumor microenvironment in IDH-mutant glioma.

BACKGROUND: Recent single-cell transcriptomic studies report that IDH-mutant gliomas share a common hierarchy of cellular phenotypes, independent of genetic subtype. However, the genetic differences between IDH-mutant glioma subtypes are prognostic, predictive of response to chemotherapy, and correlate with distinct tumor microenvironments. RESULTS: To reconcile these findings, we profile 22 human IDH-mutant gliomas using scATAC-seq and scRNA-seq. We determine the cell-type-specific differences in transcription factor expression and associated regulatory grammars between IDH-mutant glioma subtypes. We find that while IDH-mutant gliomas do share a common distribution of cell types, there are significant differences in the expression and targeting of transcription factors that regulate glial identity and cytokine elaboration. We knock out the chromatin remodeler ATRX, which suffers loss-of-function alterations in most IDH-mutant astrocytomas, in an IDH-mutant immunocompetent intracranial murine model. We find that both human ATRX-mutant gliomas and murine ATRX-knockout gliomas are more heavily infiltrated by immunosuppressive monocytic-lineage cells derived from circulation than ATRX-intact gliomas, in an IDH-mutant background. ATRX knockout in murine glioma recapitulates gene expression and open chromatin signatures that are specific to human ATRX-mutant astrocytomas, including drivers of astrocytic lineage and immune-cell chemotaxis. Through single-cell cleavage under targets and tagmentation assays and meta-analysis of public data, we show that ATRX loss leads to a global depletion in CCCTC-binding factor association with DNA, gene dysregulation along associated chromatin loops, and protection from therapy-induced senescence. CONCLUSIONS: These studies explain how IDH-mutant gliomas from different subtypes maintain distinct phenotypes and tumor microenvironments despite a common lineage hierarchy.

The evolution of alternative splicing in glioblastoma under therapy.

BACKGROUND: Alternative splicing is a rich source of tumor-specific neoantigen targets for immunotherapy. This holds promise for glioblastomas (GBMs), the most common primary tumors of the adult brain, which are resistant to standard-of-care therapy. Although most clinical trials enroll patients at recurrence, most preclinical studies have been done with specimens from primary disease. There are limited expression data from GBMs at recurrence and surprisingly little is known about the evolution of splicing patterns under therapy. RESULT: We profile 37 primary-recurrent paired human GBM specimens via RNA sequencing. We describe the landscape of alternative splicing in GBM at recurrence and contrast that to primary and non-malignant brain-tissue specimens. By screening single-cell atlases, we identify cell-type-specific splicing patterns and novel splicing events in cell-surface proteins that are suitable targets for engineered T cell therapies. We identify recurrent-specific isoforms of mitogen-activated kinase pathway genes that enhance invasiveness and are preferentially expressed by stem-like cells. CONCLUSION: These studies shed light on gene expression in recurrent GBM and identify novel targets for therapeutic development.

CpG-creating mutations are costly in many human viruses.

Mutations can occur throughout the virus genome and may be beneficial, neutral or deleterious. We are interested in mutations that yield a C next to a G, producing CpG sites. CpG sites are rare in eukaryotic and viral genomes. For the eukaryotes, it is thought that CpG sites are rare because they are prone to mutation when methylated. In viruses, we know less about why CpG sites are rare. A previous study in HIV suggested that CpG-creating transition mutations are more costly than similar non-CpG-creating mutations. To determine if this is the case in other viruses, we analyzed the allele frequencies of CpG-creating and non-CpG-creating mutations across various strains, subtypes, and genes of viruses using existing data obtained from Genbank, HIV Databases, and Virus Pathogen Resource. Our results suggest that CpG sites are indeed costly for most viruses. By understanding the cost of CpG sites, we can obtain further insights into the evolution and adaptation of viruses.

Illuminating Women's Hidden Contribution to Historical Theoretical Population Genetics.

While productivity in academia is measured through authorship, not all scientific contributors have been recognized as authors. We consider nonauthor "acknowledged programmers" (APs), who developed, ran, and sometimes analyzed the results of computer programs. We identified APs in Theoretical Population Biology articles published between 1970 and 1990, finding that APs were disproportionately women (P = 4.0 × 10-10). We note recurrent APs who contributed to several highly-cited manuscripts. The occurrence of APs decreased over time, corresponding to the masculinization of computer programming and the shift of programming responsibilities to individuals credited as authors. We conclude that, while previously overlooked, historically, women have made substantial contributions to computational biology. For a video of this abstract, see: https://vimeo.com/313424402.

The Phenotypes of Proliferating Glioblastoma Cells Reside on a Single Axis of Variation.

Although tumor-propagating cells can be derived from glioblastomas (GBM) of the proneural and mesenchymal subtypes, a glioma stem-like cell (GSC) of the classic subtype has not been identified. It is unclear whether mesenchymal GSCs (mGSC) and/or proneural GSCs (pGSC) alone are sufficient to generate the heterogeneity observed in GBM. We performed single-cell/single-nucleus RNA sequencing of 28 gliomas, and single-cell ATAC sequencing for 8 cases. We found that GBM GSCs reside on a single axis of variation, ranging from proneural to mesenchymal. In silico lineage tracing using both transcriptomics and genetics supports mGSCs as the progenitors of pGSCs. Dual inhibition of pGSC-enriched and mGSC-enriched growth and survival pathways provides a more complete treatment than combinations targeting one GSC phenotype alone. This study sheds light on a long-standing debate regarding lineage relationships among GSCs and presents a paradigm by which personalized combination therapies can be derived from single-cell RNA signatures, to overcome intratumor heterogeneity. SIGNIFICANCE: Tumor-propagating cells can be derived from mesenchymal and proneural glioblastomas. However, a stem cell of the classic subtype has yet to be demonstrated. We show that classic-subtype gliomas are comprised of proneural and mesenchymal cells. This study sheds light on a long-standing debate regarding lineage relationships between glioma cell types.See related commentary by Fine, p. 1650.This article is highlighted in the In This Issue feature, p. 1631.