ABSTRACT From publicly available next-gen sequencing datasets of non-model organisms, such as marine protists, arise opportunities to explore their evolutionary relationships. In this study we explored the effects that dataset and model selection have on the phylogenetic inference of the Gonyaulacales, single celled marine algae of the phylum Dinoflagellata with genomes that show extensive paralogy. We developed a method for identifying and extracting single copy genes from RNA-seq libraries and compared phylogenies inferred from these single copy genes with those inferred from commonly used genetic markers and phylogenetic methods. Comparison of two datasets and three different phylogenetic models showed that exclusive use of ribosomal DNA sequences, maximum likelihood and gene concatenation showed very different results to that obtained with the multi-species coalescent. The multi-species coalescent has recently been recognized as being robust to the inclusion of paralogs, including hidden paralogs present in single copy gene sets (pseudoorthologs). Comparisons of model fit strongly favored the multi-species coalescent for these data, over a concatenated alignment (single tree) model. Our findings suggest that the multi-species coalescent (inferred either via Maximum Likelihood or Bayesian Inference) should be considered for future phylogenetic studies of organisms where accurate selection of orthologs is difficult.