| The deep sequencing of second generation sequencing technology has enabled us to study complex biological structures, which have multiple DNA units simultaneously such as transcriptomics and metagenomics. Unlike general genome sequence assembly, a DNA unit of these biological structures may have multiple copies with small or substantial structural variations and/or SNPs simultaneously in an experimental sample. Therefore, the deep sequencing is necessary to figure out such variations concurrently.;This dissertation focuses on de novo transcriptome assembly which requires simultaneous assembly of multiple alternatively spliced gene transcripts. In practice, the de novo transcriptome assembly is the only option for studying the transcriptome of organisms that do not have reference genome sequences, and it can also be applied to identify novel transcripts and structural variations in the gene regions of model organisms. We propose WEAV for the de novo transcriptome assembly which consists of two separate processes: clustering and assembly.;WEAV reduces the complexity of RNA-seq dataset by partitioning it into clusters called clustering. WEAV simplify a diverse RNA-seq dataset, which has many genes together, into many, smaller clustered read sets, which have few genes a cluster, in the clustering process. The underlying idea is straightforward. A sequencer samples reads from random place so reads from one gene may have overlaps with others if sequencing depth is enough. The overlaps are the keys to connect reads from one gene. We can transform a dataset into a graph where each read is a node and two reads are connected by an edge when they have an overlap. Each connected component will be a clustered read set. As a result, we can assume that a cluster may have one or few genes; therefore, it will not be mixed.;After this process, WEAV assembles the clustered read set with de Bruijn graph backbone, and a novel error correction process simplify the backbone with a fast mapping tool, PerM. Roughly speaking, WEAV tries to solve the historical Shortest Common Superstring problem with the graph to identify multiple alternatively spliced gene transcripts simultaneously and approaches the problem using Set Cover problem. We propose novel statistical measures to make the NP hard problem manageable. The measures are explainability based on the likelihood of sequences and correctness based on bootstrapping.;We compared WEAV with other assemblers with various, simulated reads. We tested the performance by widely used measures such as specificity, sensitivity, N50, and the length of the longest sequence. After this, we tested WEAV using an experimental dataset having 58.58 million 100bp human brain transcriptome reads. WEAV assembled 156,494 contigs that were longer than 300bp. 96.3% (specificity) of these contigs were mapped onto either RefSeq, Gencode or human Genome sequences (hg19), and they covered >72% sequenced bases annotated in RefSeq and Gencode. These high sensitivity and specificity showed the exceptional power of WEAV for transcriptome assembly. |
