RNA-Seq - Science method

Hi Xingyu,

ideally RNA-seq and qPCR give you a very similar answer. You can try and double-check a couple of things: Is the qPCR optimized (product does not come up too early nor too late), are your reference genes actually stable in the conditions you test (you might want to consult the RNA-seq results), are the genes generally of low abundance (which corresponds to higher variability)?

If you still have the RNA sample you used for RNA-seq you could perform qPCR on the exact same samples.

Cheers,

Lukas

Thank you Govindkumar Balagannavar, for clearing my query. It was really helpful.

Hi Mito

you'll need some expertise in bioinformatics to get this road.

starting from fastq files, you'll need to extract data, normalize them, align them on a genome and count the transcripts...once it's done, you'll have a count matrix you'll be able to use for your BW construction.

all the best

fred

Analyzing differential gene expression in single-cell RNA sequencing (scRNA-seq) data involves several steps. Below is a generalized workflow for obtaining differentially expressed genes (DEGs) from a scRNA-seq dataset contrasting two groups:

Here's a simplified example using the Seurat package in R:

RCopy code# Install and load necessary libraries install.packages("Seurat") library(Seurat) # Load and preprocess the scRNA-seq data seurat_object <- Read10X(data.dir = "path/to/data") seurat_object <- CreateSeuratObject(counts = seurat_object) # Perform normalization, identification of highly variable genes, and dimensionality reduction seurat_object <- NormalizeData(seurat_object) seurat_object <- FindVariableFeatures(seurat_object) seurat_object <- ScaleData(seurat_object) seurat_object <- RunPCA(seurat_object) # Cluster the cells seurat_object <- FindNeighbors(seurat_object) seurat_object <- FindClusters(seurat_object) # Perform cluster differential expression analysis de_results <- FindMarkers(seurat_object, ident.1 = 1, ident.2 = 2) # Visualize the results DoHeatmap(seurat_object, features = de_results$gene, group.by = "orig.ident")

Note: This is a simplified example, and the actual analysis may vary based on the characteristics of your dataset and the tools you choose. Additionally, always consult the documentation of the specific tools you are using for detailed guidance.

I just wanna update that I have the answer to this question. Anyone interested in this question can discuss it further.

It is preferable to avoid downsampling (randomly reducing the number of reads in some samples).

Normalization techniques like DESeq2, edgeR, and limma are preferred for handling differences in sequencing depth across samples, preserving the maximum amount of information. Downsampling, on the other hand, can be used to equalize sequencing depth across all samples but may reduce statistical power, especially in samples with lower read depths. Downsampling may be considered if computational capacity is limited, data size is significant, or extreme cases have vast differences in read depths. In most cases, normalization methods are preferable.

Generally you want to treat all your samples the same way, as minor things can introduce bias (different person handling the samples, different batch of reagents, different flow cell etc). If you have any internal controls, use those for QC, if not, a housekeeping gene to check for possible biases on those 10 samples.

Thank you Mansoor Hayat for your explanation now i understand better

I think the gene ID in your GTF or GFF3 files you used for constructing the alignment index might not be the transcript ID including splice variant, which cause multiple alignment to 1 gene. I think you'd better use the genome annotation file and sequence file (gtf and fa) file from the ensembl (with gtf available at https://ftp.ensembl.org/pub/release-110/gtf/homo_sapiens/Homo_sapiens.GRCh38.110.gtf.gz and fa at https://ftp.ensembl.org/pub/release-110/fasta/homo_sapiens/dna/Homo_sapiens.GRCh38.dna.toplevel.fa.gz) or download pre-bulit index from the website or your alignment tools.

If the genome is incomplete or poorly annotated, a transcriptome-based analysis using a custom-assembled transcriptome is preferred. Hybrid approaches that combine information from both genome and transcriptome data may be considered.

In single-cell RNA sequencing (scRNA-seq) data analysis, the preprocessing steps, including denoising and normalization, are essential to ensure that the data is suitable for downstream differential expression analysis. While it may seem intuitive to directly use denoised raw data without normalization for differential analysis, this approach is generally not recommended for several reasons:

In your example with C. elegans muscle cells at different ages, it's important to normalize the data to account for differences in library size between cells. This ensures that any observed changes in gene expression are more likely to reflect true biological differences rather than technical artifacts.

That said, you should still perform differential expression analysis on the normalized data. If a gene like Cox-4 is significantly downregulated with age, the normalization should help highlight this effect by reducing the influence of technical factors.

In summary, while it may seem reasonable to directly use denoised raw data for differential analysis, it's generally recommended to perform proper normalization to account for technical variations and ensure the robustness and biological relevance of your results in scRNA-seq data analysis

General answer: Everything will affect RNA-seq results. More specific answer: compare levels of housekeeping genes between the two methodologies and see if they show any discordance. If not, you are likely good to go with your analysis

There is a similar question.

If you don't want to use sra-tools, you can search for the SRA id at https://www.ebi.ac.uk/ena/browser/ and download the fastq files directly.

samples are the key points in molecular biology and higher technologies. did you check for quality of your samples before going to library preparation. low quality (RIN) will give low target assignment.

you can also check after sequencing quality using multiQC tools but the worst is done.

fred

Here are some of them:

Hi Ahmad

before going to the software, be sure the samples were extracted with the right protocol and kits to allow miRNA to be selected or selected in the whole samples.

fred

Mehjabeen Javed I would NOT sonicate your sample, this likely sheared your RNA. try grinding with mortar and pestle with liquid nitrogen to cool the mortart and pestle, add Trizol, slowly allow to thaw, then immediately transfer to dounce homogenizer for 33 strokes. could also pass through a 31G needle if you don't have a dounce homogenizer.

Dear all,

thanks for the insightful commentaries of my colleagues. However, I beg to differ in some detail as this problem goes a bit deeper. (Just my 0.02$)

1. Throwing three samples into one is, of course, a bad idea because you waste ressources. Why not make three differently barcoded libraries and then send them for sequencing on one lane. You do not loose information and you can always ask for more reads if your sequencing depth is not sufficient. Thus you save in sequencing costs but keep all the options.

2. Do not make technical replicates. If you master the technique they will be +/- identical. If you have technical problems no replicate will help you anyway.

3. If you run biological replicates I wouldn't use the classical R-programs. Most assume that there is a "true" value that you can't measure because of random variation in your method/sample. However that is not exactly what happens in nature. Imagine you derive three transgenic cell lines with an inducible transcription factor to find target genes. Now you compare 3 times TFon/TFoff. You get following values:

TFon TFoff

geneA

sampleA: 1000 100

sampleB: 100 10

sampleC: 10 1

It is clear that geneA is very interesting. However, if you define sampleA,B,C as triplicate most analysis programs will throw this gene out because base-line expression has a higher variation as the overall difference between on/off.

Alas, geneA may be a perfect and important target gene as you do not control the overall concentration of the transcripiton factor in the transgenic cells. So it may be perfectly ok that you see this large variation in base expression. Fold-change is what counts here. In biology a value very frequently depends on more than one factor and not all can be controlled. Classical statistics fails in this cases.

Therefore I'd recommend to run barcoded libraries - evaluate each one individually and look for the intersection of genes that come up as interesting in all three instances. Then follow up on these.

In the end no statistics can replace the good old biological confirmatory experiment anyway (although "wet" biology seems to be out of fashion nowadays).

Good luck with your experiment.

Best

Robert

@ Kangkon Saikia Thanks so much for your answer. Yes, I'm using protein sequences as well. It will be really helpful if you can explain this part in your answer a bit more. "For this you have to perform gene prediction first followed by annotation of the CDS/transcripts."

Thanks again!

you can reach out to vikram@aiims.edu , they have bsl3 facilty.

Oxford Nanopore announced that they are training their base caller to recognise m1Ψ in direct RNA sequencing data at their London Calling 2023 conference. I am unsure of its current availability. You may want to ask on the Oxford Nanopore sequencing Community page for the latest on this topic.

It is better to consider batch as a factor in the design formula. The tximport pipeline proposed by Michael Love himself offers the most useful solution. Please have a look.

Chat-GPT really is a jabber-box. Context must be critically reworked by a person knowing the topic. Answers like the one posted by Rana above are not only not useful but also potentially misleading, not to say wrong. But most relevantly, it has nothing to do with a scientific interaction of researchers.

Hi Jan,

You can try PROGENY scores directly for pathways. You can get a pathway activity scores here : https://bioconductor.org/packages/release/bioc/html/progeny.html

You can also consider Gene Set Variance Analysis. https://bioconductor.org/packages/devel/bioc/vignettes/GSVA/inst/doc/GSVA.html This is can be used as a one sample GSEA and you can analyse how the samples are different in your favourite pathways (gene sets) instead of genes. You can find biological outliers or the response to each one compound.

A simple sodium acetate precipitation of the RNA will clean up your A260/230 ratios.  Here is the method I use:

- Add ultra pure water make up RNA samples to 100ul in total

- Add 300ul 100% EtOH & 10ul 3M Sodium Acetate per sample

- Vortex + incubate overnight @ -20oC

- Centrifuge full speed @ 4oC for 30mins

- Pipette off supernatant.

- Add 500ul 70% EtOH + mix

- Centrifuge full speed @ 4oC for 15-20mins

- Pipette off supernatant & air dry for at least 15 mins [check that no EtOH remaining]

- Resuspend in 30ul or 12ul (depending on RNA concentration)

- Enjoy good A260/230 values!

Give it a go, it works really well.  Just remember the RNA pellets will be difficult to see (pellet paint can help).

qPCR is considered the more sensitive and direct analysis. You use the RNA-seq to get a list of "interesting candidate genes" to then investigate using qPCR. You don't need to do any direct comparison to the original RNAseq data.

You can't know that the high TPM from RNA seq will also have high expression during qPCR. That's why you do the experiment.

Yes with the kit, Agilent, and or Qiagen,

No, kit is ok

Normalizing RNA-seq data generated from PolyA and Total RNA can be challenging, as they have different properties that can affect their library preparation and sequencing, resulting in differences in the sequencing depth and composition of the resulting reads.

One common method for normalizing RNA-seq data is to use the reads per kilobase per million mapped reads (RPKM) or transcripts per million (TPM) approach. These methods normalize for sequencing depth and gene length and provide a measure of the expression level of each gene relative to other genes within the same sample. However, RPKM and TPM normalization may not be suitable for comparing expression levels across samples with different RNA types, as they assume that the total transcriptome is being sequenced uniformly, which may not be true in the case of Total RNA.

Another method that can be used for normalizing RNA-seq data is to use the trimmed mean of M values (TMM) normalization approach, which is based on the assumption that most genes are not differentially expressed across the samples being compared. TMM normalization estimates a scaling factor for each sample based on the relative expression of a set of housekeeping genes, which are assumed to be stably expressed across all samples.

A more advanced method for normalizing RNA-seq data is to use a normalization technique that takes into account the differences between the RNA types, such as the "DESeq2" package in R, which uses a generalized linear model to estimate size factors that adjust for the differences in sequencing depth and composition between PolyA and Total RNA samples.

In summary, while RPKM or TPM normalization can be used for normalizing RNA-seq data generated from PolyA and Total RNA, it may not be suitable for comparing expression levels across different RNA types. TMM normalization can be used as a general normalization method, while advanced techniques such as DESeq2 can be used to specifically address the normalization of RNA-seq data generated from different RNA types.

Thank you, Wilhelm,

That still did not address the main question, though. Please check again.

After 2nd strand synthesis step, there is an SPIA amplification step, SPIA primer can bind to only the second strand and amplify only that strand. The first strand does not have the SPIA primer binding site, so it does not get amplified.

This may lead to a high quantity of single-stranded cDNA (due to SPIA amplification) and very low double-stranded cDNA if at all there is any.

So the question is, How did the manufacturer claim that the product will be double-stranded cDNA? whereas whatever is shown in the diagram, can generate only one strand.

as Péter Gyarmati said the extraction method might introduce an important bias in the quantification results. How did you define "differentially expressed genes"? 25% of down expression will not pass the fold-change threshold in RNA-Seq analysis. In future experiments, it could be a good idea to add a spike in RNA-Seq samples to compare the results with RT-qPCR

Honestly, DESeq2 takes almost no time to run: it's by far the easiest part of the entire pipeline.

So...do it all. All comparisons.

Male treated vs male untreated -> effects of treatment on males specifically

Female treated vs female untreated -> effects of treatment on females specifically

Male and female treated vs male and female untreated -> sex-agnostic effects of treatment

Male untreated vs female untreated -> sex-specific differences under normal conditions

Male treated vs female treated -> sex-specific differences under treated conditions

male (all) vs female (all) -> treatment-agnostic effect of sex

What you'd hope is that the same genes would crop up under the same sort of comparisons (i.e. the best hits for male treated vs male untreated would be the same as those for female treated vs female untreated), and any genes that bucked the trend by being highly altered in males but not females (or vice versa) would also pop out of the other datasets, allowing you to determine whether the differences reflect sex alone, or are a consequence of sex + treatment.

RNAseq data always gives you far, far more candidates than you can realistically handle, so you can be quite aggressive in your culling of "interesting but not that interesting" candidates, but making multiple comparisons in this manner allows you to interpret your results with more nuance. And it's really quick and easy to do.

Mitra Riasi

Many R packages are available for DEGs analysis. You may go through the following sources.

1. ccb.jhu.edu/software/stringtie/index.shtml?t=manual#deseq

2. rnabio.org/course/#module-03-expression

3. www.nature.com/articles/nprot.2012.016

4. www.nature.com/articles/nprot.2016.095

Welcome to RNA-seq! It's a crazy and wild world. You will find that responses will depend a great deal on what you're aiming to achieve. So take my responses with this in mind...it depends!

Get as close to 100% as possible otherwise you'll be having to perform a set of validation experiments to ensure that any interesting findings are due to changes in your cell type of interest and not in a "contaminating" cell type. Single cell RNA-seq may be suitable here since you'll be able to get some cell type resolution and identify the populations in which the change is occurring, of course you'll still need to validate. The strenght of scRNA-seq is that you don't need to purify/enrich your population since these get resolved as part of the procedure/analysis. However, the a drawback with scRNA-seq is that you will loose a lot of low abundant transcripts, "dropout" is also a major issue. So, if you're comfortable loosing some info on potentially valuable transcripts then scRNA-seq may be the way to go. They do potentially complement each other especially because with bulk, you may get data about low expressed transcripts. But a big caveat, it all depends! You may consider identifying and collaborating with someone with expertise in RNA-seq (sample prep and data analysis) at your local institution.

Which papers? It depends. Start with papers that are answering a similar question to yours, then dig into what would be best for you study. You can consider reaching out to representative of companies like 10X Genomics and Miltenyi Biotec...that's also a good starting point. Good luck!

Just to confirm before sending it for library prep.

I would highly recommend you that before directly jumping into using certain tools for the analysis, please try to understand the basics behind the data, data types & structures, whys and whats of data processing. And for galaxy, it is a very good platform with very good tutorials. Please go through tutorial before asking question which can easily be solved with minimal self inputs.

I am not an expert in this by any means, but I have read a lot and have seen this type of data and interpreted it before as well. I can give you what I know from my experience and others may chime in.

It is a read out of transcripts that correspond to that particular site. it might be referred to as base resolution expression of the particular sequence. Essentially the higher number of transcripts that coincide with that particular sequence the higher the score. It could be areas that were difficult to resolve due to all kinds of aspects. 1) Sequence has a lot of repeats if that was the case you would see the same resolution in the other two samples but that does not seem to be the case. These areas might be resolved better if you increase the read depth of the study.

2) It may be more suggestive of a difficulty to read them. These results may be affected by post modification of the RNA as well.

This paper describes this is clinical samples but that does not restrict the affect only to humans post-modification

Sci Adv. 2021 Aug; 7(32): eabd2605.

Published online 2021 Aug 4. doi: 10.1126/sciadv.abd2605

PMCID: PMC8336963

PMID: 34348892

Judging by the fact the title on the samples say drought I might think a more epigenetic effect (Post-modification of the RNA maybe due to stress)

I nominate DNA methylation clocks to measure aging. The problem is not that it's expensive, but that they don't work in the context of intervention. People need to double-check on whatever their methylation clock is saying with a good old life span study. But nobody has time for that anymore, and if they did it, they wouldn't like the result. Instead, they just take the clock on faith.

It is generally not recommended to use a viral RNA extraction kit for the isolation of human RNA, as the kit is specifically designed for the isolation of viral RNA and may not efficiently extract human RNA. Additionally, the reagents and protocols used in the kit may not be optimized for the isolation of human RNA and may lead to poor quality or quantity of RNA. It is recommended to use a kit specifically designed for the isolation of human RNA, such as the Quick-RNA™ MicroPrep Kit-Zymo Research.

Yes, it is possible to extract total RNA from PAXgene blood RNA tubes using the PAXgene blood RNA kit. The PAXgene blood RNA kit is specifically designed for extracting RNA from blood samples that have been collected and stored in PAXgene blood collection tubes.

To extract total RNA from a portion of the blood sample collected in a PAXgene blood RNA tube, you will need to follow the manufacturer's instructions for using the PAXgene blood RNA kit. This typically involves the following steps:

It is worth noting that the specific procedures and reagents used in the PAXgene blood RNA kit may vary depending on the specific kit that you are using. Therefore,

How does your institute financially support the applicants?

Well there is no method or procedure which can be universally used to normalize the data as you are looking for.

Normalization is done for different samples/conditions within the data and not across or between multiple datasets. And that is what the paper mentioned above is dealing with. Normalization methodology for a data set.

Further, its not just the data sets are different, they are generated by different platform. Thus, in addition to the batch effect, there can be several other factors which should also be considered before normalization.

The plot shows that the average quality per base alongside your 150 pb reads is very high which is good. This result is kind of normal when the sequencing has been outsourced. Most companies will give you prefiltered fastq files containing only reads with high quality. You can ask your sequencing provider if that was the case, although sometimes you can also find that info in the report they send together with your fastq files.

Wei Zhou

Thanks Abhijeet & Sonja.

After I deeply check the data. Forward has more over represented sequence of Illumina nextra adaptor/Index and Reverse has highly over represented by Poly (G) sequence. Finally we dropped this data and asked the vendor to redo the sequence.

In NGS, read counts have somehow to be mapped to genes. Reads are not always matching to only a unique gene, so it partially depends on the mapping algorithm how many "reads" are attributed to a gene.

Further, genes may be expressed in different variants, and the presence or abscence of an exon may impact the number of reads attributed to that gene.

For genes that are not regulated very strongly, the method used for normalization can impact the eventually observed direction of regulation.

qPCR is sensitive to the assay performance and the stability of the chosen reference genes. If amplification efficiencies are not ideal, results may change depending on the absolute Ct values of the gene. However, these are eventually "human errors" of not properly validating the qPCR assays and reference genes.

These are possible reasons that come into my mind. There may be other reasons.

U can mail me here nkrdas2@gmail.com

depending on cell types, but average amount of RNA is 2-10pg by cell...

all the best

fred

Hello Antony, which protocols have you used? In all of them you get the same result shown in the photo?

GTEX, gnomAD, COSMIC, TCGA, Protein Atlas, ensemble

The large fold-change indicates that the gene is likely not expressed to a high level. Under control conditions the experession can be almost zero, and a slightly larger expression under treatment conditions will result in a very large fold-change.

Low-expressed genes give only low or no counts in RNA seq. It might be that genes with no or very few counts are filtered out from the analysis, because the counts are not reliable. If the gene is not at all detected under control conditions (0 counts in all control samples), it is not possible to calculate any (finite) fold-change at all.

Hi Yeogha,

Firstly, what proportion of your cells do you expect to be epithelial? From my understanding of human lungs, a large proportion of the cell types are mesenchymal, immune, or endothelial - so your results might not be that unreasonable.

Secondly, have you tried any of the canonical markers of the epithelial subtypes? p63, KRT5, MUC5AC, SCGB1A1, FOXJ1 for example?

In saying that, I know very little about scRNA-seq too, so I'm sure someone else will have a much more sensible answer!

All the best,

Sam

Hi,

you need to compare the disease versus control samples when doing the statistical test. I didn't fully get if the problem is that you lack the information of which samples belongs to which category or it is a coding problem. As for the former, usually datasets have a metadata file in which the sample names you find in the gene expression table are present, and the treatment information is included. If it is a coding problem, you can index the sample names to divide the data into Asthmatic, Healthy and COPD.

You can store RNA at room temperature as an ethanol precipitate. If you have RNA in solution, take some known amount - say 10-20 ug- and add 1/10 vol. NaAcetate and 2.5 volumes ethanol. The precipitate is stable and can be mailed. Upon receiving, spin down in a microcentrifuge and resuspend the RNA pellet in a specific volume of dH2O or whatever buffer you need.

Hmm, I'm not entirely sure how to run that myself, but since I lack coding experience I've been using the LatchBio platform. Through their RNA-seq and DeSeq2 workflows, I've analyzed and visualized a lot of the data at my lab. Hope that helps!

Sheikk I share you this guide:

Is under development, but I consider that is pretty good.

Also I think you must check the quality of your assembly, and in the guide that I passed you is very clear, but you can see an expanded explanation in Trinity's Github:

Hi Abhay,

You could try to 'pseudo-bulk' your colleague's scRNA-seq data and then compare that to your own.

Here, the expression values of all profiled cells in a single-cell experiment are bulked together for each replicate. This then can be processed using the same pipeline as your bulk data giving you a much more direct comparison.

There are multiple tutorials online: https://hbctraining.github.io/scRNA-seq_online/lessons/pseudobulk_DESeq2_scrnaseq.html

One would expect concordant fold-changes to validate underlying biology (ko effect) but I wouldn't expect results to be identical though... Some differences would arise from different batches (replicates), library chemistry, sequencing runs, etc.

Also, if you still have some cells available you could run some qPCR on a few selected genes to tell you which dataset is better to trust.

I found this article quite helpful and self-explanatory for RNA seq analysis.

The following useful RG links are also very useful:

What was the N for your RNA seq, and qPCR analysis? Surprisingly, the trend is the opposite, which shouldn't be the case, unless the primer isn't specific and amplifies some other segment.

thanks Erik

Thanks Saubhik Sengupta...

Run HiSat2 on each sample individually and then combine them using samtools

Katie A S Burnette I used 3 separate cultures. The gene cluster was expressed in 2/3. But nothing was detected via MS. And yeah I have tested standards and limit of detection on the MS prior to this experiment

i am working on cardiomyopathy patients Blood samples . and wanted to do miRNA sequencing can some one please suggest how many millions reads i need to sequence 20 millions or 30 millions and also please suggest the platform as well .

How did this turn out? Curious to know if the UV light or hydrogen peroxide worked for causing DNA damage without killing Hek293 cells.

Dear Shivreet Kaur ,

In recent years, RNA sequencing (in short RNA-Seq) has become a very widely used technology to analyze the continuously changing cellular transcriptome, i.e. the set of all RNA molecules in one cell or a population of cells. One of the most common aims of RNA-Seq is the profiling of gene expression by identifying genes or molecular pathways that are differentially expressed (DE) between two or more biological conditions. This tutorial demonstrates a computational workflow for the detection of DE genes and pathways from RNA-Seq data by providing a complete analysis of an RNA-Seq experiment profiling Drosophila cells after the depletion of a regulatory gene.

Regards,

Shafagat

Hi! I recommend using LatchBio for RNA seq data. I've used it several times, and it is super easy to use since it has a non-coding interface.

Hope that helps!

Thanks, Kyle and Mohamed, for your feedback. Our RNA purifications using this kit have shown excellent yield, A260/230 and A260/280 ratios, and integrity, but they have only been used for RT-qPCR. Based on your comments, they should also be suitable for RNA-seq.

or maybe one transcript is present and the other absent, not expressed....

quality in RNAseq is not assessed by this way, fastqc is better.

Thank you Mehmet Tardu

Perhaps the LatchBio platform might be of use? It does not sequence data but it's a great RNA-seq analysis pipeline.