Assigning Fold Change for Polyester

Question

[priyankamaripuri]

Hey!

I am simulating RNA-Seq data using Polyester. My goal is to evaluate the performance of different workflows (Different combinations of tools). To simulate differentially expressed reads by Polyester, I used a Fold Change Distribution of 4-fold up-regulation (3% of transcripts), 2-fold up (7%), 1.5-fold up (9%), 1.5-fold down-regulation (6%), 2-fold down (3%), and 4-fold down(2%) and the rest 70% with a Baseline expression of 1. However, I'm very curious to know if it would make a difference if I were to take a range of Fold Change values (between 1.5 to 4). I could use the 'simulate_experiment_empirical' function from Polyester instead which simulates the reads based on a real dataset to define the counts. But, I'm concerned that it would add bias based on the tool that I use to calculate the abundances. Thanks in advance!

Answer 1

[h.mon]

Your first simulation is very artificial and I fear its conclusions would have limited relevance in evaluating the tools.

But, I'm concerned that it would add bias based on the tool that I use to calculate the abundances.

One solution is to perform several simulations, based on several tools.

Another solution is to find data sets with lots of samples, and consider the "truth" the analysis of the whole set - under the assumption all tools will converge to the truth given enough data. This is the path taken by several recent papers evaluating RNAseq analysis methods.

[priyankamaripuri]

Thank you for your reply. I am considering Simulated data for additional evaluation. I will also be using Benchmarking data and RT-qPCR info alongside the simulated data for evaluation.

One solution is to perform several simulations, based on several tools.

I have several combinations of tools and it's very tedious to make several simulations.