Video summary
GL4HS Capstone Lecture Metadata, Multi-Factor Analysis, and Statistical + Biological Significance
Main summary
Key takeaways
Main ideas / concepts covered
Course flow (Modules 2 → 3)
- Today’s focus:
- What teams need to do for Module 2
- How that work connects to Module 3
- Teams were working on team-building/networking, and should ensure:
- At least one member per team has access to the course materials.
Access troubleshooting for Google Classroom
- Instructors deleted and re-added users who couldn’t access the classroom.
- If users still see “invitation not found”, they should:
- Follow prior instructions (from Miss Clolio / “Miss Clilio had…”), including:
- Clearing browsing data
- Trying different browsers
- Follow prior instructions (from Miss Clolio / “Miss Clilio had…”), including:
- If problems persist, students should contact instructors via:
- Slack, or
- A message containing “here.”
Module 1 → Module 2 (Why metadata matters)
- Module 1 is a deep dive into metadata, not a quick look.
- Students must research:
- What tissues were investigated
- What is unknown
- Research directions
- Papers already derived from the dataset
- Purpose:
- Avoid proposing work someone else already did
- Build enough background to form a strong hypothesis for a research proposal
Why standardized preprocessing/pipelines matter
- Data were pre-processed by a gene lab using a standardized RNA-seq pipeline for:
- Quality control
- Reproducibility
- Key QC and normalization concepts:
- RNA quality and assay quality determine data validity
- The pipeline includes stopping points with quality control
- Spike-ins (known RNA quantity/quality) help normalize data and detect library-prep inefficiencies
- Enables cross-dataset comparison (common in open data repositories)
End-to-end analysis pipeline (what to verify before analysis)
The pipeline includes:
- Hypothesis generation / experimental design
- RNA purification + library preparation
- Sequencing
- Quality control
- Trimming
- Alignment
- Quantification
- Differential gene expression
- Biological insight + visualization
Critical point:
- Students must understand not only what was planned, but what actually happened.
- Example: differences between animal cohorts due to unexpected events (e.g., a natural disaster / ground control changes).
Detailed methodology / step-by-step instructions (as presented)
A) Use OSDR metadata thoroughly (Module 1 emphasis)
In the OSDR experiment description, check:
- Mission/acronym meaning (e.g., rodent research reference missions)
- Mouse strain (varies in traits like bone loss susceptibility or immunocompromise)
- Age (interpret in model-organism terms; mouse “old” ≈ 24 months+)
Use factors from metadata because they drive later notebook configurations.
Then click into metadata-linked items:
- Organism pages
- General biology + sequencing context
- Project/study page
- Mission background, center, duration in space, strain
- Payload/protocol sections
- Time points, baseline timing, housing density changes, supplements, cage conditions
Also note potential confounders that could explain unexpected genes:
- Housing density differences
- Seasickness/rough-conditions stress during spaceflight
- Radiation details (type/dose; often shielded, but still potentially influential)
Review protocol sections for:
- Sample collection + tissue preservation
- Transcription protocols
- Library construction
- Sequencing method
- Data processing steps
B) Quality control workflow using MultiQC reports (Module 2 emphasis)
In OSDR files/notebooks, use gene-lab processed RNA-seq outputs to find:
- MultiQC “merge sequence data” quality control reports (HTML)
Steps:
- Select/download the multiqc report files for the dataset
- Open the downloaded HTML to view plots covering multiple samples:
- Flight live/return samples vs ground controls vs young/old subsets
QC checks highlighted:
- Read length distribution should match expected ~150
- GC content should be tight/narrow
- Per-base quality scores:
- Expected to stay in the “green” region
- Drop-off toward later positions can be normal
- Adapter content:
- Assess whether trimming is needed
- Metrics can be visualized by configuring/plotting columns and exporting plots
C) Trim-related QC (pre vs post trimming)
After QC, verify trimming by:
- Returning to OSDR protocol sections to see trimming parameters
- Confirming pipeline details, such as:
- Raw FASTQ filtered using Trim Galore powered by cutadapt
- FastQC then MultiQC generation post-filtering
- Trimming parameters used for the alignment-ready dataset
Compare:
- Raw vs trimmed QC MultiQC reports
- Changes in GC content / read length / % failed to estimate trim impact
D) Alignment and quantification (overview-level)
- Alignment:
- Follow the standardized pipeline parameters used by the gene lab team
- In notebooks, alignment parameters are assumed (not emphasized as user-configurable)
- Quantification:
- Used to identify expressed genes and downstream expression changes
E) Multifactors + differential expression notebooks (Capstone notebooks)
- Notebooks (e.g., notebook 6/7 in the basic course, and capstone notebooks 1/2) are dataset-specific.
- Students must ensure:
- The notebook’s OSD dataset number is correct
- Chosen configurations match the dataset’s factor values in metadata
Recommended starting configuration:
- Basic microgravity vs 1G comparison
- Spaceflight vs ground control (or parabolic flight vs ground control, depending on dataset)
Other metadata factors that may exist:
- Age (young vs old)
- 1G by centrifugation (1G condition in space vs ground)
- Live animal return vs terminated on ISS
- Possibly sex, genotype, duration/time in space
Configuration workflow:
- Copy notebooks and re-run for different factor-value comparisons
- Use the notebook’s “define factors/configuration” sections
- Optionally run multiple configurations across teammates, then compile results
Important caution:
- Factor values must match metadata exactly (string matching is strict: spaces, capitalization, exact genotype names)
- Units might appear on the website but not inside notebook factor/value lists—use notebook-listed values.
How to compare combined factors (multi-factor design):
- Define a factor list with multiple factors
- Create groups such as:
- Group A = young + ground control
- Group B = old + ground control
This enables comparisons like:
- genotype effects under flight vs ground
- age × condition combinations
Gene lists feed into:
- Pathway analysis tools
- Gene Set Enrichment Analysis (GSEA)
- Over-representation analysis (including GO/ORA)
Instructor note (e.g., Dr. Calto / James):
- Prefilled configs may work for some notebooks but not others.
- One config (e.g., “config 5” type behavior) is guaranteed to work because its output was verified by executing the notebooks.
Statistics + significance + biological significance (conceptual lessons)
Statistical significance (p-values)
- p-value = probability of observing the data if the null hypothesis is true (no real difference).
- RNA-seq involves multiple testing:
- Thousands of genes tested (e.g., ~30,000)
- Even p = 0.01 can produce many false positives when repeated across tests
Therefore outputs include:
- Adjusted p-values using multiple hypothesis correction:
- Bonferroni correction
- FDR / Benjamini-Hochberg-style correction
Biological significance
- Statistical significance alone is not enough.
- Biological significance considers:
- Fold change
- Whether the change is meaningful beyond natural variation
Fold change context examples:
- Strongly repressed genes:
- Small fold changes can still be biologically meaningful
- Post-transcriptional regulation:
- Large fold changes may not translate to meaningful biological impact
Pathways vs individual genes
- Pathway analysis can use less stringent thresholds because:
- Pathways assess co-expression across genes already significant in gene-level analysis.
- Students must still check biological plausibility:
- Avoid importing genes into pathway analysis that aren’t actually significant in the dataset, or results may be misleading.
ANOVA vs t-test
- t-test: compares two groups
- ANOVA: compares more than two groups
- Post hoc comparisons can follow ANOVA to inspect pairwise differences
Students’ Q&A themes included in the lecture
- Whether recordings will be posted:
- Yes, but may take time to process
- How publications relate to dataset pages:
- Publications using the dataset are flagged/linked (possibly with a lag)
- Students can search by dataset number in publications/methods
- Strategies for literature reviews in groups:
- Instructors plan to share literature review templates in Google Classroom
- Clarification on “significance”:
- Distinguishes statistical vs biological significance (especially under multiple testing)
- Pathway analysis vs single-gene focus:
- Recommended to do both in parallel:
- pathway analysis for context/mechanistic interpretation
- individual gene inspection for biomarkers/targeted insights
- Recommended to do both in parallel:
- Clarification on “custom” pathway options:
- Pathways may come from external/custom datasets/databases, not only built-in pathways
Next steps / logistics
- A second part of pathway analysis will be in a separate lecture scheduled for Monday (per instructor plan).
- Notebooks:
- Will be shared in Slack (not Google Classroom)
- Instructor intends to release notebooks over the weekend if possible
- Event reminders:
- Last organized Kumo space at 4 PM (timezone noted)
- Guest lecture by “Twimman Clemens” (CEO of Space Tango) at 12:00 PM Pacific time
- Admin:
- Slides and recording will be posted when available
Speakers / sources featured (identified)
Speakers
- Liz (instructor/presenter; primary voice guiding content on OSDR/QC/statistics)
- Dr. Calto (co-instructor; clarifications on notebook configs and multi-factor setup)
- James (co-instructor; notebook configuration examples and factor-value matching)
- Sahi (student asking/engaging in Q&A)
- Jennifer (student asking about significance/pathway focus)
- Hashita (student asking about literature review strategies)
- Anvi/Anvie (student asking about focusing on single genes vs pathway analysis, and other questions)
Other referenced people/sources
- Miss Clolio / Miss Clauddio (referenced for Classroom access troubleshooting)
- Dr. Alaska (example of person/center affiliation in OSDR metadata)
- Twimman Clemens (guest lecture speaker; CEO of Space Tango)
- Space Tango
- OSDR / Open Science Data Repository
- Gene lab sample processing team
- Tools/terms referenced:
- RNA-seq pipeline, RNA purification + library prep, ERCC spike-ins
- MultiQC, Trim Galore (cutadapt), FastQC
- STAR, GSEA, Over-representation analysis
- Gene Ontology (GO)
- Benjamini-Hochberg / FDR, Bonferroni
- ANOVA, post hoc analysis