What's your recommended length and depth for my RNA-seq experiment

[biotech analytics]

My favorite plant species has a 800 Mb genome and around 100 - 150 Mb transcriptom. It only has 15 genome assembly for 15 cultivars and several de novo RNA assembly. My cultivar has no existed genome assembly or transcriptom assembly. The published journal papers usually generates 50 Mb to 100 Mb reads per sample for de novo RNA assembly. I have a litmited budget.

1. So should I go shallow and generate 20 - 30 million reads per sample and do mapping using the existed genome assembly or transcriptom assembly of other cultivars? In this way, I can sequence more samples.
2. Should I go deep and generate above 50 Mb reads per sample an assemble my own transcriptom? In this way, I can only sequence fewer samples.
3. We are doing pilot experiments with limited budgets. My collaborator wants to have no replicate so that we can include more treatments. Is it a big no no? I personally prefer at least 3 biological replicates per each treatment.

Thank you so much for your kind help!

[GenomicSeq]

Hello biotech analytics. I think that going shallow and using an existing genome assembly or transcriptome assembly from other cultivars can be a cost-effective approach, especially if you have limited resources. This should allow you to sequence more samples within your budget. But I would also like to say there are a few caveats to consider.

The first is sequence similarity. If your cultivar is closely related to the cultivars with existing assemblies, using those references may yield good enough results. But on the other hand, if there are significant genetic differences between your cultivar and the reference genomes/transcriptomes, you can encounter challenges in accurately mapping the reads or identifying novel genes and transcripts. Also, you have to think about the biological variation. Different cultivars may have variations in their gene content and expression profiles. Using a reference assembly from a different cultivar may not fully capture the genetic and transcriptomic characteristics of your specific cultivar. This could limit your ability to detect cultivar-specific variations and gene expression patterns.

For your second question about going deep. Well, I believe that going deep and generating more reads per sample to assemble your own transcriptome can provide you with a more complete and accurate representation of your cultivar's transcriptome. This way will allow you to identify novel genes and transcripts specific to your cultivar. But again, this comes with other trade-offs, like budget limitations and assembly challenges. In general, you should consider the balance between depth and sample size based on your research goals and available resources.


And for your last question on no replicates...
I always say that having biological replicates is considered essential for robust statistical analysis and reliable interpretation of results. It's the best way to account for biological variation and assess the consistency of your findings. I get that your collaborator wants to include more treatments, but it's highly recommended to prioritize a reasonable number of biological replicates over an extensive number of treatments, especially in the context of a pilot experiment. Otherwise, it will be really hard to assess what you've done because insignificant variation will throw off your assessment.

Ideally, you should aim for a minimum of three biological replicates per treatment. It's definitely important to strike a balance between the number of treatments and the number of replicates to ensure the validity and reliability of your results.

Sorry, I know that was a lot so I'll try to summarize my points here.

1. I would prioritize a reasonable number of biological replicates per treatment (e.g., three replicates).
2. I would evaluate the degree of genetic similarity between your cultivar and existing assemblies before deciding whether to use them as references or perform de novo assembly.
3. You should find a balance between depth and sample size based on your research goals and available resources.

These are just my opinions so it wouldn't hurt to ask some of your colleagues as well on their recommendations.