Alignment-free

Large-scale comparison of the similarities between two biological sequences is a major issue in computational biology; a fast method, the D 2 statistic, relies on the comparison of the ktuple content for both sequences. Although it has been known for some years that the D 2 statistic is not suitable for this task, as it tends to be dominated by single-sequence noise, to date no suitable adjustments have been proposed. In this article, we suggest two new variants of the D 2 word count statistic, which we call D S 2 and D * 2. For D S 2 , which is a selfstandardized statistic, we show that the statistic is asymptotically normally distributed, when sequence lengths tend to infinity, and not dominated by the noise in the individual sequences. The second statistic, D Ã 2 , outperforms D S 2 in terms of power for detecting the relatedness between the two sequences in our examples; but although it is straightforward to simulate from the asymptotic distribution of D Ã 2 , we cannot provide a closed form for power calculations.

This research describes a novel, alignment-free method of genomic sequence comparisons based on absent nucleotide words and expression levels. Testing this method on Influenza A virus isolates, three classifications are presented which successfully identify; 1) the geographic origins of domestic bird H5N1 isolates through China and Southeast Asia during 2006, 2) the country of human H5N1 isolates crossing over from domestic bird hosts and, 3) the historical flu season from which human H3N2 isolates originated. Because comparison methods used do not rely on alignment, they are computationally efficient and well suited for large numbers of sequences in compehensive flu transmission network delineation.

Background: The development of high-throughput sequencing technologies and, as its result, the production of huge volumes of genomic data, has accelerated biological and medical research and discovery. Study on genomic rearrangements is crucial owing to their role in chromosomal evolution, genetic disorders, and cancer. Results: We present Smash++, an alignment-free and memory-efficient tool to find and visualize small-and large-scale genomic rearrangements between 2 DNA sequences. This computational solution extracts information contents of the 2 sequences, exploiting a data compression technique to find rearrangements. We also present Smash++ visualizer, a tool that allows the visualization of the detected rearrangements along with their self-and relative complexity, by generating an SVG (Scalable Vector Graphics) image. Conclusions: Tested on several synthetic and real DNA sequences from bacteria, fungi, Aves, and Mammalia, the proposed tool was able to accurately find genomic rearrangements. The detected regions were in accordance with previous studies, which took alignment-based approaches or performed FISH (fluorescence in situ hybridization) analysis. The maximum peak memory usage among all experiments was ∼1 GB, which makes Smash++ feasible to run on present-day standard computers.

CHAPTER 4. Application of the STing algorithm TO public health and environmental genomics 70 4.1 Applying STing to public health: Shiga toxin-producing Escherichia coli (E. coli) virulence profiling 4.1.1 Materials and methods 4.1.2 Results and discussion 4.1 Applying STing to environmental genomics: nifH gene-based taxonomic assignment of amplicon sequencing samples 4.1.1 Materials and methods 4.1.2 Results and discussion CHAPTER 5. Conclusions and future prospects 102 PUBLICATIONS 107 APPENDIX A. SUPPLEMENTARY DATA FOR CHAPTER 2 108 A.1 Pseudocode for database indexing A.2 Pseudocode for sequence typing APPENDIX B. SUPPLEMENTARY DATA FOR CHAPTER 3 173 B.1 WebSTing data dictionary APPENDIX C. SUPPLEMENTARY DATA FOR CHAPTER 4 177 xiv SUMMARY Public health agencies increasingly couple next generation sequencing (NGS) based characterization of microbial genomes with bioinformatics analysis methods for molecular epidemiology. The overhead associated with the bioinformatics methods used for this purpose, in terms of both the required human expertise and computational resources, represents a critical bottleneck that limits the potential impact of microbial genomics on public health. This is particularly true for local public health agency laboratories, which are typically staffed with microbiologists who may not have substantial bioinformatics expertise or ready access to high-performance computational resources. There is a pressing need for bioinformatics solutions to genome-enabled molecular epidemiology that is accurate, easy to use, fast, and computationally efficient. The development of an alignment-free algorithm for NGS data analysis and its implementation into turn-key software applications tailored explicitly for genome-enabled molecular epidemiology and environmental microbial genomics is the focus of my research. I explored a computational strategy based on k-mer frequencies to distinguish among sequences of interest in NGS read samples. By combining this strategy with the efficient data structure enhanced suffix array (ESA), I developed a base algorithm for the rapid analysis of unprocessed NGS reads. I further adapted and implemented this algorithm into a suite of software applications for sequence typing, gene detection, and gene-based taxonomic read classification. My thesis research focused on three specific aims: (1) development of an alignment-and assembly-free algorithm and software solution for NGS-based molecular xv epidemiology, (2) development of an alignment-and assembly-free fully automated Webplatform for the comprehensive characterization of bacterial isolates using whole genome sequencing (WGS) data; and (3) expanding the applicability of the alignment-free algorithm to different problems. Sequence typing and gene detection are essential for pathogen characterization in genome-enabled approaches to molecular epidemiology. In this sense, I developed an assembly-and alignment-free algorithm, STing, which I implemented into two turn-key software utilities for sequence typing, and gene detection. Benchmarking and validation analyses showed that STing is an ultrafast and accurate solution for genome-enabled molecular epidemiology, which performs better than existing bioinformatics methods for sequence typing and gene detection. Limited access to bioinformatics-related infrastructure and expertise impedes the successful adoption of genome-enabled approaches to molecular epidemiology in public health. To overcome this challenge, I developed WebSTing, a Web-platform that uses the STing algorithm to supply easy access to the accurate and rapid alignment-free automated characterization of WGS samples of bacterial isolates. To demonstrate the utility of STing in problems beyond simple sequence typing and gene detection, I applied the alignment-free algorithm to two different areas: (1) public health, with the virulence gene profiling of Shiga toxin-producing Escherichia coli (STEC) isolates, and (2) environmental microbial genomics, with the nifH gene-based taxonomic classification of amplicon sequencing reads. I showed that STing performs better than the xvi gold-standard method for STEC isolate characterization and that it correctly classifies amplicon sequencing reads on simulated communities of nitrogen-fixing organisms. Research advance 1: A novel k-mer frequencies approach combined with the ESA data structure was used to develop an assembly-and alignment-free algorithm, STing, for rapid analysis of unprocessed NGS reads. The STing algorithm was implemented into two software applications for genome-enabled molecular epidemiology: (1) the STing typer for sequence typing, and (2) the STing detector for gene detection. The STing typer utility was compared to six widely used programs for genome-enabled sequence typing, using the traditional multilocus sequence typing (MLST) scheme, and two larger typing schemes, ribosomal MLST (rMLST), and core genome MLST (cgMLST). Comparison results showed that STing outperformed the other applications in terms of accuracy and efficiency (runtime and RAM) using the MLST and rMLST schemes and was second with the cgMLST scheme. Most importantly, STing was the only application able to perform the typing analysis using all three of the typing schemes assessed, while also showing the ability to scale successfully to genome-enabled typing schemes like cgMLST. The detector utility was used to evaluate the ability of STing for detecting two epidemiologically relevant types of markers: antimicrobial resistance (AMR) genes and virulence factor (VF) genes. Results showed that STing had 100% accuracy in detecting AMR and VF genes on 71 WGS samples generated from 17 bacterial species of high priority in clinical microbiology research. Research advance 2: A Web-based platform, WebSTing, was developed to provide fully automated NGS-based characterization of bacterial pathogens. The main goal of WebSTing to provide easy access to genome-enabled approaches to molecular xvii epidemiology in public health laboratories that have limitations in bioinformatics-related infrastructure and expertise. WebSTing uses the STing algorithm and supplies assemblyand alignment-free sequence typing, gene detection, and phylogenetic analysis of WGS samples of bacterial isolates. Research advance 3: The applicability of the STing algorithm was expanded to solve problems in two different areas: (1) public health, and (2) environmental microbial genomics. For the public health application, STing was used for virulence gene profiling STEC isolates from WGS samples. Here, STing was compared to the PCR method, the current gold-standard technique used by public health laboratories for STEC characterization. Results showed that STing is more accurate than PCR for characterizing virulence genes in STEC samples. STing showed between 98% and 100% accuracy in characterizing the four genes used as markers for STEC determination (stx1, stx2, eae, and ehxA), compared to a PCR accuracy between 90% and 94%. Most importantly, and unlike the PCR technique, STing was able to detect novel genes in the analyzed STEC isolates. In the environmental microbial application, the STing algorithm was extended for nifH gene-based taxonomic classification of amplicon sequencing reads. The algorithm was implemented into the STing classifier utility. Using a nifH gene reference database, the STing classifier program was able to classify reads correctly in samples of nitrogen-fixing organism communities, simulated up to the lowest sequencing depth of 1x coverage. Importantly, results showed that full-length reference sequences of the gene nifH led to better taxonomic classification of the sequencing reads.

In a genomic sequence, the oligonucleotide signature represents the ratio of the observed to expected number of occurrences of all possible nucleotide words of a specific length. Word absence is also found in genomic sequences whereby specific words are not observed despite their expected presence in a random nucleotide distribution. This research uses a combination of word absence and oligonucleotide signatures to quantify differences between inter-epidemic and intra-epidemic Influenza A virus (H3N2) genomic sequences. In addition, word absence/presence patterns are examined for their discrimination of sequences from distinct epidemics or geographic origins within the same epidemic. Interepidemic sequences are well delineated by word absences and difference measures into epidemic specific groups. Intra-epidemic sequences are not consistently well separable in terms of their geographic origins, but show similarities across geographic regions.

With the progress of modern sequencing technologies a large number of complete genomes are now available. Traditionally the comparison of two related
genomes is carried out by sequence alignment. There are cases where these
techniques cannot be applied, for example if two genomes do not share the
same set of genes, or if they are not alignable to each other due to low sequence similarity, rearrangements and inversions, or more specifically to their
lengths when the organisms belong to different species. For these cases the
comparison of complete genomes can be carried out only with ad hoc methods
that are usually called alignment-free methods.

The understanding of the whole human genome and of other species, and
the mechanisms behind replication and evolution of genomes are some of the
major problems in genomics. Although most of the current methods in genome
sequence analysis are based only on genetic and annotated regions, this could
saturate the problem because of their limited size of information. In fact,
recent evidence suggests that the evolutionary information is also carried by
the non-genic regions [35], and in some cases we cannot even estimate a complete phylogeny by analyzing the genes shared by a clade of species [19].
Accordingly, this chapter addresses the phylogeny reconstruction problem for
different organisms, namely viruses, prokaryotes, and unicellular eukaryotes,
using whole-genome pairwise sequence comparison.