Current best practice for biomarker discovery using transcriptomics data

Question

[Javad]

Hi Guys,

I am about to start working on a biomarker discovery project using transcriptomics and I basically have no experience in biomarker discovery (although I have experience with rna-seq data analysis).

Also from theoretical point of view, I am informed of how biomarker discovery using rna-seq works.

My problem is that I don’t know the state of the art bioinformatics tools that are being used in this field. Therefore, I looked for a tutorial or a step by step workflow that I can use to quickly catch up the current best practice for transcriptomics based biomarker Discovery but I can’t find a good one.

I really appreciate if someone can introduce a good tutorial or a step by step workflow.

Thanks in advance

Answer 1

[thyleal]

I never came across such tutorial...

If you are familiar with RNAseq data analysis for differential expression, you can follow from there.

What I often see is:

• Exploratory data analysis and quality control (PCA, MDS, sanity checks etc);
• Differential gene/transcript analysis (interesting to select the most differentially expressed genes in average, not necessarily the most discriminant ones);
• Depending on your sample size and what type of dependend variable you have (categorical two level, categorical multilevel, ordinal, numeric, etc.), you can train some models, e.g, penalized regression (lasso, ridge, elastic net), random forests, support vector machines, kNN etc;
• Validate on external datasets (public) or in new samples (or test set from your own experiment, not used in any steps before).

There are more things to do, but depends on study design, data quality, biological phenomenon and sample size and study objectives.

I suggest start reading good papers from reputable journals with similar study design. Also, there are some good reviews on transcriptomic biomarkers and rnaseq biomarkers or for classification purposes.

[Javad]

Thanks.

I am also familiar with the machine learning steps and this part is not that much of an issue.

Can you introduce few recent reviews that you mentioned?

[thyleal]

Sure, take a look:

• http://dx.doi.org/10.1016/j.ymeth.2012.08.012
• 10.3389/fgene.2020.00019
• 10.1186/s13045-020-01005-x
• 10.2217/bmm.15.16
• https://rna-seqblog.com/tag/biomarker-discovery/ (good collection of posts)
• 10.3389/fimmu.2019.00221
• 10.6026/97320630014153

If I remember more I'll update here.

[Javad]

One more point is that the biomarkers are not limited to DE. For example SNV, aberrant splicing, fusion, etc. can also have diagnostic value. If you can elaborate on these aspects, it would also be very valuable.

[thyleal]

Yes, I thought you were only interested in expression biomarkers (mRNA, lncRNA, miRNA etc.).