Individual differences among deep neural network models, bioRxiv, 2020-01-09
AbstractDeep neural networks (DNNs) excel at visual recognition tasks and are increasingly used as a modelling framework for neural computations in the primate brain. However, each DNN instance, just like each individual brain, has a unique connectivity and representational profile. Here, we investigate individual differences among DNN instances that arise from varying only the random initialization of the network weights. Using representational similarity analysis, we demonstrate that this minimal change in initial conditions prior to training leads to substantial differences in intermediate and higher-level network representations, despite achieving indistinguishable network-level classification performance. We locate the origins of the effects in an under-constrained alignment of category exemplars, rather than a misalignment of category centroids. Furthermore, while network regularization can increase the consistency of learned representations, considerable differences remain. These results suggest that computational neuroscientists working with DNNs should base their inferences on multiple networks instances instead of single off-the-shelf networks.
biorxiv neuroscience 100-200-users 2020Three Rules Explain Transgenerational Small RNA Inheritance in C. elegans, bioRxiv, 2020-01-09
Life experiences trigger transgenerational small RNA-based responses in C. elegans nematodes. Dedicated machinery ensures that heritable effects would re-set, typically after a few generations. Here we show that isogenic individuals differ dramatically in the persistence of transgenerational responses. By examining lineages composed of >20,000 worms we reveal 3 inheritance rules (1) Once a response is initiated, each isogenic mother stochastically assumes an “inheritance state”, establishing a commitment that determines the fate of the inheritance. (2) The response that each mother transfers is uniform in each generation of her descendants. (3) The likelihood that an RNAi response would transmit to the progeny increases the more generations the response lasts, according to a “hot hand” principle. Mechanistically, the different parental “inheritance states” correspond to global changes in the expression levels of endogenous small RNAs, immune response genes, and targets of the conserved transcription factor HSF-1. We show that these rules predict the descendants’ developmental rate and resistance to stress.
biorxiv genetics 100-200-users 2020A simple, cost-effective, and robust method for rRNA depletion in RNA-sequencing studies, bioRxiv, 2020-01-07
AbstractThe profiling of gene expression by RNA-sequencing (RNA-seq) has enabled powerful studies of global transcriptional patterns in all organisms, including bacteria. Because the vast majority of RNA in bacteria is ribosomal RNA (rRNA), it is standard practice to deplete the rRNA from a total RNA sample such that the reads in an RNA-seq experiment derive predominantly from mRNA. One of the most commonly used commercial kits for rRNA depletion, the Ribo-Zero kit from Illumina, was recently discontinued. Here, we report the development a simple, cost-effective, and robust method for depleting rRNA that can be easily implemented by any lab or facility. We first developed an algorithm for designing biotinylated oligonucleotides that will hybridize tightly and specifically to the 23S, 16S, and 5S rRNAs from any species of interest. Precipitation of these oligonucleotides bound to rRNA by magnetic streptavidin beads then depletes rRNA from a complex, total RNA sample such that ~75-80% of reads in a typical RNA-seq experiment derive from mRNA. Importantly, we demonstrate a high correlation of RNA abundance or fold-change measurements in RNA-seq experiments between our method and the previously available Ribo-Zero kit. Complete details on the methodology are provided, including open-source software for designing oligonucleotides optimized for any bacterial species or metagenomic sample of interest.ImportanceThe ability to examine global patterns of gene expression in microbes through RNA-sequencing has fundamentally transformed microbiology. However, RNA-seq depends critically on the removal of ribosomal RNA from total RNA samples. Otherwise, rRNA would comprise upwards of 90% of the reads in a typical RNA-seq experiment, limiting the reads coming from messenger RNA or requiring high total read depth. A commonly used, kit for rRNA subtraction from Illumina was recently discontinued. Here, we report the development of a ‘do-it-yourself’ kit for rapid, cost-effective, and robust depletion of rRNA from total RNA. We present an algorithm for designing biotinylated oligonucleotides that will hybridize to the rRNAs from a target set of species. We then demonstrate that the designed oligos enable sufficient rRNA depletion to produce RNA-seq data with 75-80% of reads comming from mRNA. The methodology presented should enable RNA-seq studies on any species or metagenomic sample of interest.
biorxiv microbiology 200-500-users 2020Engineering E. coli for magnetic control and the spatial localization of functions, bioRxiv, 2020-01-07
AbstractThe fast-developing field of synthetic biology enables broad applications of programmed microorganisms including the development of whole-cell biosensors, delivery vehicles for therapeutics, or diagnostic agents. However, the lack of spatial control required for localizing microbial functions could limit their use and induce their dilution leading to ineffective action or dissemination. To overcome this limitation, the integration of magnetic properties into living systems enables a contact-less and orthogonal method for spatiotemporal control. Here, we generated a magnetic-sensing Escherichia coli by driving the formation of iron-rich bodies into bacteria. We found that these bacteria could be spatially controlled by magnetic forces and sustained cell growth and division, by transmitting asymmetrically their magnetic properties to one daughter cell. We combined the spatial control of bacteria with genetically encoded-adhesion properties to achieve the magnetic capture of specific target bacteria as well as the spatial modulation of human cell invasions.
biorxiv synthetic-biology 0-100-users 2020Glacier ice archives fifteen-thousand-year-old viruses, bioRxiv, 2020-01-07
AbstractWhile glacier ice cores provide climate information over tens to hundreds of thousands of years, study of microbes is challenged by ultra-low-biomass conditions, and virtually nothing is known about co-occurring viruses. Here we establish ultra-clean microbial and viral sampling procedures and apply them to two ice cores from the Guliya ice cap (northwestern Tibetan Plateau, China) to study these archived communities. This method reduced intentionally contaminating bacterial, viral, and free DNA to background levels in artificial-ice-core control experiments, and was then applied to two authentic ice cores to profile their microbes and viruses. The microbes differed significantly across the two ice cores, presumably representing the very different climate conditions at the time of deposition that is similar to findings in other cores. Separately, viral particle enrichment and ultra-low-input quantitative viral metagenomic sequencing from ∼520 and ∼15,000 years old ice revealed 33 viral populations (i.e., species-level designations) that represented four known genera and likely 28 novel viral genera (assessed by gene-sharing networks). In silico host predictions linked 18 of the 33 viral populations to co-occurring abundant bacteria, including Methylobacterium, Sphingomonas, and Janthinobacterium, indicating that viruses infected several abundant microbial groups. Depth-specific viral communities were observed, presumably reflecting differences in the environmental conditions among the ice samples at the time of deposition. Together, these experiments establish a clean procedure for studying microbial and viral communities in low-biomass glacier ice and provide baseline information for glacier viruses, some of which appear to be associated with the dominant microbes in these ecosystems.ImportanceThis study establishes ultra-clean microbial and viral sampling procedures for glacier ice, which complements prior in silico decontamination methods and expands, for the first time, the clean procedures to viruses. Application of these methods to glacier ice confirmed prior common microbiological findings for a new ice core climate record, and provides a first window into viral genomes and their ecology from glacier ice across two time horizons, and emphasizes their likely impact on abundant microbial groups. Together these efforts provide clean sampling approaches and foundational datasets that should enable simultaneous access to an archived virosphere in glacier ice.
biorxiv ecology 100-200-users 2020Single-cell epigenomic identification of inherited risk loci in Alzheimer’s and Parkinson’s disease, bioRxiv, 2020-01-07
ABSTRACTGenome-wide association studies (GWAS) have identified thousands of variants associated with disease phenotypes. However, the majority of these variants do not alter coding sequences, making it difficult to assign their function. To this end, we present a multi-omic epigenetic atlas of the adult human brain through profiling of the chromatin accessibility landscapes and three-dimensional chromatin interactions of seven brain regions across a cohort of 39 cognitively healthy individuals. Single-cell chromatin accessibility profiling of 70,631 cells from six of these brain regions identifies 24 distinct cell clusters and 359,022 cell type-specific regulatory elements, capturing the regulatory diversity of the adult brain. We develop a machine learning classifier to integrate this multi-omic framework and predict dozens of functional single nucleotide polymorphisms (SNPs), nominating gene and cellular targets for previously orphaned GWAS loci. These predictions both inform well-studied disease-relevant genes, such as BIN1 in microglia for Alzheimer’s disease (AD) and reveal novel gene-disease associations, such as STAB1 in microglia and MAL in oligodendrocytes for Parkinson’s disease (PD). Moreover, we dissect the complex inverted haplotype of the MAPT (encoding tau) PD risk locus, identifying ectopic enhancer-gene contacts in neurons that increase MAPT expression and may mediate this disease association. This work greatly expands our understanding of inherited variation in AD and PD and provides a roadmap for the epigenomic dissection of noncoding regulatory variation in disease.
biorxiv genomics 100-200-users 2020