In the analysis of any data using statistical modelling, it is imperative that the choice of model is informed by expert knowledge and that its adequacy is determined based on the extent to which it captures and describes the patterns observed in the data. This is especially true in systems where a subset of the constituent components may not be known or cannot be observed. In this chapter, we demonstrate how statistical inference can be used to inform model selection and, by identifying where existing models are unable to sufficiently capture observed behaviour, that statistical inference can help indicate which model refinements may be required.

In this chapter, we use Bayesian statistical methodology – specifically, Riemannian manifold population MCMC – to model interactions between molecular species in the JAK/STAT pathway in chronic myeloid leukaemia (CML) and compare two candidate models. We set out the biological context for this inference in Sections 9.1–9.1.4 and describe the two candidate models in Section 9.3. With the biology established, we describe our statistical methodology (Section 9.4) which we successfully apply in a simulation study to provide a proof of concept (Section 9.5), before we consider a subsequent, more biologically realistic dataset (Section 9.6) to assess which model best describes the behaviour observed in vitro. We relate the findings from this second synthetic study back to our model and dataset construction, thereby highlighting what further in vitro and in silico work is required (Section 9.7).

The oncology of chronic myeloid leukaemia

The condition that we now recognise as chronic myeloid leukaemia (CML) was first described in 1845, in quick succession, by two pathologists, Dr John Hughes Bennett (Bennett 1845) and Dr Rudolf Virchow (Virchow 1845).

Rapid advances in genetics, genomics and imaging have given insight into the molecular and cellular basis of behaviour in a variety of model organisms with unprecedented detail and scope. It is increasingly becoming routine to isolate behavioural mutants, clone and characterise mutant genes and discern the molecular and neural basis for a behavioural phenotype. Conversely, reverse genetic approaches have made it possible to straightforwardly identify genes of interest in whole-genome sequences and generate mutants that can be subjected to phenotypic analysis. In this latter approach, it is the phenol typing that presents the major bottleneck; when it comes to connecting phenotype to genotype in freely behaving animals, analysis of behaviour itself remains superficial and time-consuming. However, many proof-of-principle studies of automated behavioural analysis over the last decade have poised the field on the verge of exciting developments that promise to begin closing this gap.

In the broadest sense, our goal in this chapter is to explore what we can learn about the genes involved in neural function by carefully observing behaviour. This approach is rooted in model organism genetics but shares ideas with ethology and neuroscience, as well as computer vision and bioinformatics. After introducing Caenorhabditis elegans as a model, we will survey the research that has led to the current state of the art in worm behavioural phenol typing and present current research that is transforming our approach to behavioural genetics.

The worm as a model organism

Caenorhabditis elegans is a nematode worm that lives in bacteria-rich environments such as rotting fruit and has also been isolated from insects and snails which it is thought to use for longer-range transportation (Barriere & Felix 2005, Lee et al. 2011). In the laboratory, it is commonly cultured on the surface of agar plates seeded with a lawn of the bacterium Escherichia coli as a food source. On plates, worms lie on either their left or right side and crawl by propagating a sinuous dorso-ventral wave from head to tail.

Despite immense achievements in the past century in hygiene control, and the development of vaccines and antibiotics, infectious diseases continue to pose a tremendous threat to public health globally. There are still devastating infections for which there are no effective vaccines or antimicrobial therapies. Moreover, the problem of drug resistance in bacteria and viral populations and the increasing appreciation that pathologies resulting from infections are responsible for a number of chronic conditions, are creating an ever-growing need for novel preventive and therapeutic approaches. In line with this, a new host-targeted approach has been suggested for antimicrobial drug research that exploits the central role of the host cell during infection. Decades of research have taught us that infections are supported by host cell functions, and that infection pathology is frequently host dependent. Accordingly, the pharmacological targeting of host cell factors promises novel opportunities to prevent and treat infectious disease. Such an approach may be anticipated to expand the number of druggable targets, produce broad-spectrum compounds and impede the generation of resistance. The discovery of RNA interference (RNAi) has created opportunities to explore gene functions in cellular systems in a targeted manner. RNAi loss-of-function approaches have proved invaluable for the identification of host proteins important for pathogen viability. These approaches can be applied on a high-throughput scale, which demands sophisticated liquid handling and high-content image analysis. Here, we provide an overview of the current status of high-content screening (HCS) in loss-of-function analyses in infectious disease research and discuss how these powerful techniques can be applied to identify host factors with previously unknown roles in infection and its pathology.

The challenge of fighting infectious diseases

Infections by pathogenic species of bacteria, viruses, fungi and protozoa have had considerable impact on mankind throughout history. Advances in our understanding of the importance of hygiene control, and later, improvements in diagnostics and the development and successful employment of vaccines and antimicrobial drugs, have substantially benefited human health, and provided social and economic benefits.

Despite great strides in the development of anti-cancer strategies over the last 50 years, treatment regimens continue to cause significant toxicity and fail to fully eradicate disease. Enhancing the current state of therapy will require: (1) the expansion of available tumor selective and therapeutically tractable molecular targets, (2) the development of methods to provide a rational approach to identifying effective combinatorial drug cocktails, and (3) molecular markers that can accurately predict sensitive patient populations. To this end, efforts that reveal the molecular architecture supporting tumorigenic phenotypes are essential. RNA interference (RNAi)-mediated loss of function screens have emerged as a method for wholesale identification of tumor-specific dependencies that modulate chemo responsiveness. Here, we provide a broad overview of how genome-scale RNAi screening is being implemented.

Cancer chemotherapy

Cytotoxic chemotherapy

Goodman and Gilman's 1946 discovery that lymphosarcomas respond to nitrogen mustard demonstrated that tumor cells may have an enhanced sensitivity to chemical poisons as compared to their normal counterparts. This finding revolutionized cancer treatment as it indicated that in addition to radiation and surgery, the only available modalities at the time, drugs could also be administered to reduce tumor burden (Goodman et al. 1946). Following on these initial observations, over the ensuing 50 years, an arsenal of cytotoxic agents were developed to treat a range of cancer types (Chabner & Roberts 2005, Strebhardt & Ullrich 2008).

The majority of these agents, as with the nitrogen mustard, share a common characteristic: they induce genomic damage. For example, agents such as cisplatin cause inter-and intrastrand DNA cross links. This DNA damage can lead to the inhibition of cell division by activating an arrest in the cell cycle to allow for DNA repair through the nucleotide excision repair (NER) pathway. This pathway is coupled to apoptotic programs that are activated if overwhelming damage is detected (Plunkett et al. 1995, Siddik 2003, Wang & Lippard 2005).

How to link genotypes and phenotypes is a long-standing question in modern biology. Modern high-throughput approaches are key technologies at the forefront of genetic research. They enable the analysis of a biological response to thousands of experimental perturbations and require a tight collaboration between experimental and computational scientists. Perturbation studies and computational approaches have revolutionized research in functional genomics and genetics and promise to lay the foundation for personalized medicine. For modern high-throughput technologies, computation is as important as experimentation. Genome-wide image-based RNA interference (RNAi) screens, for example, are only feasible because of computational techniques. Computational skills to analyse the data have become as important as experimental skills to generate the data.

Design and analysis of phenol typing screens depend on the number of genes perturbed and the richness of the phenotype observed (Figure 6.1). At one extreme are high-throughput screens with single reporters, e.g. a genome-wide screen for new components of a pathway. At the other extreme are perturbations of individual genes with very rich phenotypes, e.g. assessing the effects of a single gene perturbation on several molecular levels over time. Between these two extremes lie a variety of possible screen designs. Two widely used scenarios are small-scale perturbations (<20 genes) of a single target pathway with rich readouts, e.g. a global transcriptional profile, and medium-scale perturbations (hundreds of genes) with multi-parametric readouts, e.g. cell morphology or growth in different media. In the following we will discuss statistical and computational methodologies for functional analysis in all four scenarios.

Scenario 1: Genome-wide screens with single reporters

RNAi screens have been frequently and successfully applied for functional profiling of genes on a large scale (Boutros & Ahringer 2008). The vast majority of these applications use a single phenotype (e.g. cell viability, growth rate, activity of reporter constructs) to characterize the function of genes in specific biological pathways.

The advent of sequencing technologies has revolutionized our understanding and approach to studying biological systems. Indeed, whole-genome sequencing projects have already targeted many different species, enabling the identification of most genes in those organisms. However, observed phenotypes cannot be explained by genes alone, but rather by the interactions that their products establish under some environmental conditions (Waddington 1957). Thus, it is through the analysis of these interaction net-works (e.g. regulatory, metabolic, molecular, or genetic) that we can better understand the genotype-to-phenotype relationship, the complexity and evolution of organisms, or the differences among individuals of the same species. The topology and dynamics of these biological networks can be unveiled by systematic perturbation of their nodes (i.e. genes). For instance, upon single-gene deletions in Saccharomyces cerevisiae under standard laboratory conditions, most genes (∼80%) were not found to be essential for cell viability (Giaever et al. 2002). Though many of these genes may be required for growth in other environments (Hillenmeyer et al. 2008), this result suggests extensive functional redundancy among genes. Such functional buffering confers robustness to biological networks and shields the cellular machinery from genetic perturbations (Hartman et al. 2001). Additionally, the small effect on phenotype that many gene deletions exhibit (see Figure 2.1) evidences that single perturbations alone cannot capture the complexity of the genotype-to-phenotype relationship. Therefore, a combinatorial approach to gene perturbations is best suited to elucidate biological systems and can enable a better characterization of genes and cellular functioning.

Definition of genetic interaction

Genetic interactions reveal functional relations between genes that contribute to a pheno-typic trait. William Bateson first introduced the term, formerly known as epistasis (see Phillips [1998] for a description on the origin and evolution of the definition), to refer to an allele at one locus preventing a variant at another from manifesting its effect (Bateson 1909).