A Study of How Read Depth and Seedling Number Influence the Development of de novo Transcriptomes for Two Leucaena Species

As next generation sequencing becomes ever cheaper, efficient, and available, more research is able to focus on non-model organisms. The relatively low cost and focus on functional elements of the genome have resulted in transcriptome studies becoming popular for research on non-model organisms. However, the optimization of sample selection and read-depth has not been standardized, yet they have a profound impact on the resulting transcriptomes. Here we present data on two different out-crossing, non-model species, Leucaena trichandra and L. cruziana, belonging to the Leguminosae family. We generated Illumina libraries and HiSeq 100bp paired-end reads for a range of seedling numbers, which were used to assemble de novo transcriptomes. The results suggest that transcriptome assembly from a single seedling is not as comprehensive as those for higher order seedling number from the same maternal line. However, no significant difference is found among transcriptomes based on 3, 5, and 8 seedlings. We also explore the number of reads necessary to generate a comprehensive transcriptome assembly by subsampling the available reads. We found that 37M read-pairs are sufficient to reconstruct the more complex transcriptome, L. cruziana. Restricting the read-pairs to 31M or 24M decreases the number of transcripts recovered and projected gene number, as well as gene completeness as demonstrated by full-length orthologs found in the TRAPID database. In conclusion, we offer some guidance for other researchers who wish to assemble de novo transcriptomes in non-model plants, both by presenting our method for determining best parameters and by suggesting a starting experiment.