Purpose

In this project you will analyze an RNAseq experiment without knowing which study it comes from. Working blind lets you form your own interpretation of the data before you see what the authors concluded. The goal is not only to run the pipeline correctly, but to evaluate what the data can and cannot tell you, commit to an interpretation, and then compare it critically with the published work.

This structure is not meant as a gotcha or to make the project harder. I want you to explore the data and form your own conclusions without bias, rather than deferring to what the paper stated was “interesting”. Remember that every scientific paper is a highly curated story of what the authors found significant or novel based on their own research questions, interests, and the constraints of publication; there may be other interesting results or findings that were not as remarked upon. You are not expected to find something novel or something the authors missed; matching their findings is just as valid an outcome.

Project structure

This project has two phases:

  • Phase 1 (blind analysis): You will receive the data and sample metadata only. You will process the data, evaluate its quality, perform differential expression and enrichment analyses, and write your own interpretation of what happened in this experiment.
  • Phase 2 (informed analysis): Roughly halfway into the project (2 weeks in), I will tell you the publication and experiment this data originated from. You will then replicate figures from the paper and compare your findings with the authors’. You should not revise your Phase 1 interpretation; it’s simply meant as a way for you to compare results.

Working in an R Markdown notebook

For this project, since we are working with DESeq2, it will be easier to work in an R Markdown notebook. R Markdown notebooks are highly similar to Jupyter notebooks and allow you to write R code and markdown in the same file.

Phase 1: Blind analysis

Experimental Design (1 paragraph)

Using only the sample metadata you were given, describe the experiment:

  • How many groups are there, and how many replicates per group?
  • What covariates are provided (e.g. sex, batch, sequencing run)?
  • What comparison(s) can this design support? What kinds of effects might it be unable to detect?
  • What potential confounders would you be concerned about, and can you account for them with the information you have?

Methods (as long as needed)

Please view the presentation here for guidelines: methods

The methods section should concisely describe which steps were taken in the analysis of the data. Remember to adhere to our conventions for writing a methods section, including specifying the software versions used and the parameters for each step. If no parameters are changed, state that default parameters were used. Remember that a methods section should include any details that are necessary for someone else to replicate your analysis exactly. You can exclude any details that should not affect the results, such as the exact file paths used.

  • Write a methods section for the pipeline you developed for project 2

Read Quality Control (1-2 paragraphs)

Use the MultiQC report or the individual logs from FastQC and STAR to evaluate the quality of the reads. Please make sure that at minimum you specifically mention the following (even if there’s no issue, state that there is no issue):

  • The range of the number of reads in all the samples
  • Any potential issues with the reads as flagged by FastQC
  • Any overrepresented sequences, or adapter contamination in the reads
  • The alignment rate of the reads to the reference genome
  • Multimapping rate
  • Based on your evaluation above, please state whether you believe the experiment was of high quality and was suitable for downstream analysis. If not, please state what you would do to improve the quality of the reads.

Filtering the Counts Matrix (1-2 paragraphs)

Please describe the filtering process you used to filter the counts matrix.

  • Ensure that you include the two plots that show the distribution of counts before and after filtering
  • State the number of genes present before and after filtering
  • Justify in a paragraph why you chose to filter in this way

Sample Quality Control (2 paragraphs)

Evaluate your samples before performing differential expression. These plots tell you whether your downstream results can be trusted.

  • Follow the instructions in the DESeq2 vignette, and normalize the filtered counts by using the rlog transformation or the variance-stabilizing transformation
  • Perform PCA on the normalized counts matrix and overlay the sample information in a biplot of PC1 vs. PC2
  • Create a heatmap or graphic of the sample-to-sample distances for the experiment
  • Do the samples cluster by experimental group? Are any samples outliers? Do any covariates appear to explain the variation better than the experimental groups?
  • Based only on these plots, would you proceed to differential expression with all samples as they are? If not, what would you change, and why?

Differential Expression Analysis (3-4 paragraphs)

  • Create a table of the top 10 differentially expressed genes and the statistics provided by DESeq2 as ranked by adjusted p-value
  • Choose a padj threshold and report the number of significant genes at this threshold, separated into up- and downregulated
  • Use the thresholded results to perform a DAVID or Enrichr analysis on the significant genes at your chosen padj threshold
    • State the background gene set you used and justify it
    • State whether you submitted up- and downregulated genes together or separately, and why
  • Perform a GSEA analysis using fgsea on your RNAseq results using the C2 canonical pathways MSigDB dataset and log2 fold change as the ranking metric
  • Choose a padj threshold for the fgsea analysis and create a plot of your choice that displays the top most significantly enriched pathways
  • Comment briefly on the results of the DAVID or Enrichr analysis and the table you created and what it indicates about the biological processes that might differ between groups
  • Comment briefly on the results of the fgsea analysis by stating what you believe are the main significant pathways of interest and why?
  • Compare the results from your DAVID/Enrichr analysis and your fgsea analysis and comment on any similarities or differences you observe

Interpretations (2-3 paragraphs)

Using the number and direction of your DE genes, along with your pathway and enrichment results, please do the following:

  • Speculate broadly on the major changes caused by the experimental condition
  • Which results most strongly support your speculation?

Your Conclusions and Follow-up

  • Propose two to three additional experiments you might do to validate your findings. You do not have to do them, just propose them and anticipate what they would show you if your findings held, and what result would convince you that your findings were wrong.

  • Propose at least two follow-up experiments that would extend or expand on your findings

Phase 2: Informed analysis

Introduction (1 paragraph)

  • What is the biological background of the study?
  • Why was the study performed?
  • Why did the authors use the bioinformatic techniques they did?

Replicate Figures 3C and 3F (3-4 paragraphs)

  • Create a volcano plot similar to the one seen in figure 3C
  • Use your DAVID/Enrichr or fgsea results and create a plot resembling figure 3F but with your findings. You do not need to use the same pathways as they did.
  • Read their discussion of their results and specifically address the following in your provided notebook:

  1. Compare how many significant genes are up- and downregulated in their findings and yours (using their significance threshold). Ensure you list how many you find vs. how many they report.

  2. Compare their enrichment results with your DAVID/Enrichr and fgsea analyses. Comment on any differences you observe.

  3. List the candidate sources of discrepancy between your results and theirs (e.g. genome and annotation versions, aligner, filtering, normalization, thresholds, multiple testing correction). Using the paper’s methods as evidence, comment on at least two candidate sources you believe contribute the most to the differences you observe.

    • For each candidate source, please provide a follow-up analysis or experiment that would reveal whether it was the likely source of the differences. You do not need to actually do this analysis, just propose how.

Comparing Interpretations (2-3 paragraphs)

Return to your first interpretations about the major biological pathways implicated in the results:

  • Did the authors emphasize any genes or pathways that were weak or absent in your results? How strongly do you think their data support those claims?

  • Which of your findings agree with the authors’? Does reaching the same result through a different pipeline make you more or less confident in it, and why?

Conclusions (1-2 paragraphs)

Please address the following:

  • Compare your first conclusions with those made by the paper. Did they propose any follow-up experiments or perform any validation? What was different between what you proposed and what they did or proposed in the text?

Updated: