The deep coverage offered by next-generation sequencing (NGS) technology has facilitated the reconstruction of intra-host RNA viral populations at an unprecedented level of detail. However, NGS data requires sophisticated analysis dealing with millions of error-prone short reads. This dissertation will first review the challenges and methods for viral NGS genomic data analysis in the NGS era. Second, it presents a software tool CliqueSNV for inferring viral quasispecies based on extracting pairs of statistically linked mutations from noisy reads, which effectively reduces sequencing noise and enables identifying minority haplotypes with a frequency below the sequencing error rate. Finally, the dissertation describes algorithms VOICE and MinDistB for inference of relatedness between viral samples, identification of transmission clusters, and sources of infection.