This article discusses the importance of using transcript-level abundance estimates to improve gene-level inferences in RNA-seq data analysis. The authors argue that gene-level abundance estimates are more accurate and robust than transcript-level estimates, leading to better statistical performance and interpretability. They also show that differential isoform usage can lead to inflated false discovery rates in differential gene expression analyses, but this can be addressed by incorporating offsets derived from transcript-level abundance estimates. The study provides an R package, tximport, to help users integrate transcript-level abundance estimates into count-based statistical inference engines. The authors tested their methods on simulated and real data sets, demonstrating that incorporating transcript-level estimates improves the accuracy of differential gene expression analysis, particularly for genes with differential isoform usage. The study highlights the importance of carefully defining the research question before selecting a statistical approach and suggests that gene-level analysis may be more appropriate for many applications. The authors also note that while transcript-level analysis can be useful in certain cases, it may not always be necessary, as differential gene expression and differential transcript usage can provide sufficient information. The study concludes that using transcript-level estimates can lead to more accurate and stable statistical analysis, and that the tximport package provides a valuable tool for integrating transcript-level data into gene-level analyses.This article discusses the importance of using transcript-level abundance estimates to improve gene-level inferences in RNA-seq data analysis. The authors argue that gene-level abundance estimates are more accurate and robust than transcript-level estimates, leading to better statistical performance and interpretability. They also show that differential isoform usage can lead to inflated false discovery rates in differential gene expression analyses, but this can be addressed by incorporating offsets derived from transcript-level abundance estimates. The study provides an R package, tximport, to help users integrate transcript-level abundance estimates into count-based statistical inference engines. The authors tested their methods on simulated and real data sets, demonstrating that incorporating transcript-level estimates improves the accuracy of differential gene expression analysis, particularly for genes with differential isoform usage. The study highlights the importance of carefully defining the research question before selecting a statistical approach and suggests that gene-level analysis may be more appropriate for many applications. The authors also note that while transcript-level analysis can be useful in certain cases, it may not always be necessary, as differential gene expression and differential transcript usage can provide sufficient information. The study concludes that using transcript-level estimates can lead to more accurate and stable statistical analysis, and that the tximport package provides a valuable tool for integrating transcript-level data into gene-level analyses.