Sensitivity of transcriptomics: Different samples and methodology alter conclusions in Gulf pipefish (Syngnathus scovelli)

Abstract

Transcriptome analysis has become a central tool in evolutionary and functional genomics. However, variation among biological samples and analysis techniques can greatly influence results, potentially compromising insights into the phenomenon under study. Here, we evaluate differences in the brain transcriptome between female and male Gulf pipefish (Syngnathus scovelli). We perform comparisons between results from entire pipelines for brain transcriptome assembly, quantification, and analysis. We also offer a unique biological comparison between two sampling instances (Redfish Bay: n = 15, Port Lavaca: n = 7). Our results demonstrate crucial shortcomings with current experimental approaches. We found high variation within our results that was driven by both technical differences between pipelines and biological differences between pipefish samples. In our analysis of highly expressed genes, we found that the choice of methods influenced the degree of contamination or noise included in the identified genes. Notably, genes identified within the same pipeline were more similar than any other comparison. Our differential expression analysis revealed that both methodology and sampling location influenced the quantity and consistency of statistically significant transcripts. In the context of these results, we offer modifications to current practices that may increase the robustness of transcriptome-based conclusions. In particular, the use of a reference-guided assembly and an increase in sample sizes are likely to improve resistance to noise or error.