Machine Learning is the study of methods for programming computers to learn. Computers are applied to a wide range of tasks, and for most of these it is relatively easy for programmers to design and implement the necessary software. However, there are many tasks for which this is difficult or impossible. These can be divided into four general categories. First, there are problems for which there exist no human experts. For example, in modern automated manufacturing facilities, there is a need to predict machine failures before they occur by analyzing sensor readings. Because the machines are new, there are no human experts who can be interviewed by a programmer to provide the knowledge necessary to build a computer system. A machine learning system can study recorded data and subsequent machine failures and learn prediction rules. Second, there are problems where human experts exist, but where they are unable to explain their expertise. This is the case in many perceptual tasks, such as speech recognition, hand-writing recognition, and natural language understanding. Virtually all humans exhibit expert-level abilities on these tasks, but none of them can describe the detailed steps that they follow as they perform them. Fortunately, humans can provide machines with examples of the inputs and correct outputs for these tasks, so machine learning algorithms can learn to map the inputs to the outputs. Third, there are problems where phenomena are changing rapidly. In finance, for example, people would like to predict the future behavior of the stock market, of consumer purchases, or of exchange rates. These behaviors change frequently, so that even if a programmer could construct a good predictive computer program, it would need to be rewritten frequently. A learning program can relieve the programmer of this burden by constantly modifying and tuning a set of learned prediction rules. Fourth, there are applications that need to be customized for each computer user separately. Consider, for example, a program to filter unwanted electronic mail messages. Different users will need different filters. It is unreasonable to expect each user to program his or her own rules, and it is infeasible to provide every user with a software engineer to keep the rules up-to-date. A machine learning system can learn which mail messages the user rejects and maintain the filtering rules automatically. Machine learning addresses many of the same research questions as the fields of statistics, data mining, and psychology, but with differences of emphasis. Statistics focuses on understanding the phenomena that have generated the data, often with the goal of testing different hypotheses about those phenomena. Data mining seeks to find patterns in the data that are understandable by people. Psychological studies of human learning aspire to understand the mechanisms underlying the various learning behaviors exhibited by people (concept learning, skill acquisition, strategy change, etc.).

Machine learning

Proteins and their functional interactions form the backbone of the cellular machinery. Their connectivity network needs to be considered for the full understanding of biological phenomena, but the available information on protein-protein associations is incomplete and exhibits varying levels of annotation granularity and reliability. The STRING database aims to collect, score and integrate all publicly available sources of protein-protein interaction information, and to complement these with computational predictions. Its goal is to achieve a comprehensive and objective global network, including direct (physical) as well as indirect (functional) interactions. The latest version of STRING (11.0) more than doubles the number of organisms it covers, to 5090. The most important new feature is an option to upload entire, genome-wide datasets as input, allowing users to visualize subsets as interaction networks and to perform gene-set enrichment analysis on the entire input. For the enrichment analysis, STRING implements well-known classification systems such as Gene Ontology and KEGG, but also offers additional, new classification systems based on high-throughput text-mining as well as on a hierarchical clustering of the association network itself. The STRING resource is available online at https://string-db.org/.

https://www.zora.uzh.ch/id/eprint/165500/1/30476243.pdf

STRING v11: protein-protein association networks with increased coverage, supporting functional discovery in genome-wide experimental datasets.

Targeted nucleases are powerful tools for mediating genome alteration with high precision. The RNA-guided Cas9 nuclease from the microbial clustered regularly interspaced short palindromic repeats (CRISPR) adaptive immune system can be used to facilitate efficient genome engineering in eukaryotic cells by simply specifying a 20-nt targeting sequence within its guide RNA. Here we describe a set of tools for Cas9-mediated genome editing via nonhomologous end joining (NHEJ) or homology-directed repair (HDR) in mammalian cells, as well as generation of modified cell lines for downstream functional studies. To minimize off-target cleavage, we further describe a double-nicking strategy using the Cas9 nickase mutant with paired guide RNAs. This protocol provides experimentally derived guidelines for the selection of target sites, evaluation of cleavage efficiency and analysis of off-target activity. Beginning with target design, gene modifications can be achieved within as little as 1-2 weeks, and modified clonal cell lines can be derived within 2-3 weeks.

/pdf/genome-engineering-using-the-crispr-cas9-system-4x8nw03mgu.pdf

Genome engineering using the CRISPR-Cas9 system

Bacteria and archaea have evolved adaptive immune defenses, termed clustered regularly interspaced short palindromic repeats (CRISPR)/CRISPR-associated (Cas) systems, that use short RNA to direct degradation of foreign nucleic acids. Here, we engineer the type II bacterial CRISPR system to function with custom guide RNA (gRNA) in human cells. For the endogenous AAVS1 locus, we obtained targeting rates of 10 to 25% in 293T cells, 13 to 8% in K562 cells, and 2 to 4% in induced pluripotent stem cells. We show that this process relies on CRISPR components; is sequence-specific; and, upon simultaneous introduction of multiple gRNAs, can effect multiplex editing of target loci. We also compute a genome-wide resource of ~190 K unique gRNAs targeting ~40.5% of human exons. Our results establish an RNA-guided editing tool for facile, robust, and multiplexable human genome engineering.

/pdf/rna-guided-human-genome-engineering-via-cas9-4eefu2kibr.pdf

RNA-Guided Human Genome Engineering via Cas9

The advent of facile genome engineering using the bacterial RNA-guided CRISPR-Cas9 system in animals and plants is transforming biology. We review the history of CRISPR (clustered regularly interspaced palindromic repeat) biology from its initial discovery through the elucidation of the CRISPR-Cas9 enzyme mechanism, which has set the stage for remarkable developments using this technology to modify, regulate, or mark genomic loci in a wide variety of cells and organisms from all three domains of life. These results highlight a new era in which genomic manipulation is no longer a bottleneck to experiments, paving the way toward fundamental discoveries in biology, with applications in all branches of biotechnology, as well as strategies for human therapeutics.

The new frontier of genome engineering with CRISPR-Cas9

The application relates to a deep scanning mutagenesis library to interrogate phenotypi c changes in a population of cells comprising a plurality of CRISPR-Cas system guide RNAs targeting genomic sequences within at least one continuous genomic region, wherein the guide RNAs target at least 100 genomic sequences upstream of a PAM sequence for every 1000 base pairs within the continuous genomic region and methods for their use.

Functional genomics using crispr-cas systems for saturating mutagenesis of non-coding elements, compositions, methods, libraries and applications thereof

Provided herein are nucleic acid molecules that target the BCL11A enhancer functional regions, compositions comprising the nucleic acid molecules and methods for increasing fetal hemoglobin levels in a cell by disrupting BCL11A expression at the genomic level. Also provided herein are methods and compositions relating to the treatment of hemoglobinopathies by reinduction of fetal hemoglobin levels. In particular, the nucleic acid molecules target the +62, +58, and/or the +55 enhancer functional regions.

Targeting bcl11a enhancer functional regions for fetal hemoglobin reinduction

Abstract Immune checkpoint blockade (ICB) has transformed the treatment of metastatic cancer but is hindered by variable response rates. A key unmet need is the identification of biomarkers that predict treatment response. To address this, we analyzed six whole exome sequencing cohorts with matched disease outcomes to identify genes and pathways predictive of ICB response. To increase detection power, we focus on genes and pathways that are significantly mutated following correction for epigenetic, replication timing, and sequence-based covariates. Using this technique, we identify several genes ( BCLAF1, KRAS, BRAF , and TP53) and pathways (MAPK signaling, p53 associated, and immunomodulatory) as predictors of ICB response and develop the Cancer Immunotherapy Response CLassifiEr (CIRCLE). Compared to tumor mutational burden alone, CIRCLE led to superior prediction of ICB response with a 10.5% increase in sensitivity and a 11% increase in specificity. We envision that CIRCLE and more broadly the analysis of recurrently mutated cancer genes will pave the way for better prognostic tools for cancer immunotherapy. 

/pdf/recurrent-somatic-mutations-as-predictors-of-immunotherapy-11e25cso.pdf

Recurrent somatic mutations as predictors of immunotherapy response

A key limitation of the commonly-used CRISPR enzyme S. pyogenes Cas9 is the strict requirement of an NGG protospacer-adjacent motif (PAM) at the target site, which reduces the number of accessible genomic loci. This constraint can be limiting for genome editing applications that require precise Cas9 positioning. Recently, two Cas9 variants with a relaxed PAM requirement (NG) have been developed (xCas9 and Cas9-NG) but their activity has been measured at only a small number of endogenous sites. Here we devised a high-throughput Cas9 pooled competition screen to compare the performance of both PAM-flexible Cas9 variants and wild-type Cas9 at thousands of genomic loci and across 3 modalities (gene knock-out, transcriptional activation and suppression). We show that PAM flexibility comes at a substantial cost of decreased DNA targeting and cutting. Of the PAM-flexible variants, we found that Cas9-NG outperforms xCas9 regardless of genome engineering modality or PAM. Finally, we combined xCas9 mutations with those of Cas9-NG, creating a stronger transcriptional modulator than existing PAM-flexible Cas9 variants.

https://www.biorxiv.org/content/biorxiv/early/2020/01/23/2020.01.22.916064.full.pdf

High-throughput screens of PAM-flexible Cas9 variants for gene knock-out and transcriptional modulation

Pooled CRISPR screens have been used to identify genes responsible for specific phenotypes and diseases, and, more recently, to connect genetic perturbations with multi-dimensional gene expression profiles. Here, we describe a method to link genome-wide chromatin accessibility to genetic perturbations in single cells. This scalable, cost-effective method combines pooled CRISPR perturbations with a single-cell combinatorial indexing assay for transposase-accessible chromatin (CRISPR-sciATAC). Using a human and mouse species-mixing experiment, we show that CRISPR-sciATAC separates single cells with a low doublet rate. Then, in human myelogenous leukemia cells, we apply CRISPR-sciATAC to target 21 chromatin-related genes that are frequently mutated in cancer and 84 subunits and cofactors of chromatin remodeling complexes, generating chromatin accessibility data for ~30,000 single cells. Using this large-scale atlas, we correlate loss of specific chromatin remodelers with changes in accessibility — globally and at the binding sites of individual transcription factors. For example, we show that loss of the H3K27 methyltransferase EZH2 leads to increased accessibility at heterochromatic regions involved in embryonic development and triggers expression of multiple genes in the HOXA and HOXD clusters. At a subset of regulatory sites, we also analyze dynamic changes in nucleosome spacing upon loss of chromatin remodelers. CRISPR-sciATAC is a high-throughput, low-cost single-cell method that can be applied broadly to study the role of genetic perturbations on chromatin in normal and disease states.

https://biorxiv.org/content/10.1101/2020.11.20.390971v1.full.pdf

Neville E. Sanjana

Papers

Functional genomics using crispr-cas systems for saturating mutagenesis of non-coding elements, compositions, methods, libraries and applications thereof

Targeting bcl11a enhancer functional regions for fetal hemoglobin reinduction

Recurrent somatic mutations as predictors of immunotherapy response

High-throughput screens of PAM-flexible Cas9 variants for gene knock-out and transcriptional modulation

Scalable pooled CRISPR screens with single-cell chromatin accessibility profiling