3. Data Collection and Analysis

Learning Objectives

• Classify the major data types generated in biotechnology research: genomic, proteomic, metabolomic, transcriptomic, and structural
• Explain how experimental design choices affect the quality and reliability of collected data
• Describe the main data-generation technologies: high-throughput sequencing, microarrays, mass spectrometry, and flow cytometry
• Outline the core data analysis techniques: sequence alignment, genome assembly, and differential expression analysis
• Trace a complete study from hypothesis through data collection, analysis, and validation

Quick Answer

Data collection and analysis form the practical engine of biotechnology research — collection is how you generate or obtain the raw measurements, and analysis is how you turn those measurements into a defensible conclusion. Biotechnology deals with several distinct data types (genomic, proteomic, transcriptomic, metabolomic), each produced by specific technologies like next-generation sequencing, mass spectrometry, or flow cytometry, and each requiring matched analysis tools like sequence aligners, statistical tests, or differential expression software. Getting this stage right matters because the same raw dataset can support a valid conclusion or a misleading one depending entirely on how carefully it was collected and analyzed.

Types of Data in Biotechnology Research

Not all biological data is the same, and the type of data determines both how it's collected and how it must be analyzed:

• Genomic data — DNA sequences, gene variants, chromosomal structure
• Transcriptomic data — mRNA expression levels, alternative splicing, non-coding RNA
• Proteomic data — protein structures, post-translational modifications, protein-protein interactions
• Metabolomic data — metabolite concentrations and metabolic pathway flux
• Epigenetic data — DNA methylation patterns, histone modifications
• Structural data — 3D protein structures and ligand-binding sites

A researcher studying drought resistance in wheat, for example, might need transcriptomic data (which genes turn on under stress) alongside metabolomic data (which stress-related compounds accumulate) to build a complete picture — no single data type tells the whole story.

Why It Matters

Choosing the wrong data type for your question wastes resources. If a research question is about protein folding, generating RNA-seq data (transcriptomic) won't answer it — you need structural or proteomic data instead.

Data Collection: Experimental Design and Technologies

Experimental Design Comes First

Before any wet-lab or computational data generation begins, the underlying experimental design (randomized controlled trials, case-control studies, cohort studies, cross-sectional studies) determines what the collected data can actually prove. A dataset collected without a matched control group cannot later be fixed by clever analysis.

High-Throughput Sequencing

Next-generation sequencing (NGS) platforms — Illumina, PacBio, Oxford Nanopore — generate massive volumes of DNA/RNA sequence data quickly and relatively cheaply. Specialized variants include single-cell RNA sequencing (scRNA-seq), which captures gene expression one cell at a time instead of averaging across a tissue, and ChIP-seq, which maps where specific proteins bind DNA (used for epigenetic profiling).

Microarray Technology

Older than NGS but still used for targeted, cost-effective measurements: cDNA microarrays measure gene expression, SNP arrays detect genetic variation, and tiling arrays map genome-wide transcription. Microarrays are cheaper and simpler than NGS but only measure predefined probes, giving lower resolution and missing anything not already anticipated in the array's design.

Mass Spectrometry

The workhorse for proteomic and metabolomic data: liquid chromatography-MS (LC-MS) and gas chromatography-MS (GC-MS) separate and identify molecules by mass, while tandem MS (MS/MS) fragments molecules further for precise identification — essential for identifying which proteins or metabolites are present and in what quantity.

Flow Cytometry

Analyzes individual cells suspended in fluid as they pass through a laser, measuring size, granularity, and fluorescence. Fluorescence-activated cell sorting (FACS) goes further by physically sorting cells into separate populations based on these measurements — critical for immunology and stem cell research.

Real-World Example

To test whether climate change alters drought-gene expression in plants, researchers might expose three plant species to a randomized controlled trial (control vs. warmer temperature), collect leaf samples after 30 days, perform RNA-seq (data collection), align reads with a tool like STAR and run differential expression with DESeq2 (data analysis), then confirm the top hits using qRT-PCR (validation) — a full loop from design through validation.

Data Analysis Techniques

Sequence Alignment

Comparing DNA or protein sequences to find similarities and differences — the foundation of nearly all downstream genomic analysis. Global alignment (Needleman-Wunsch) compares sequences end-to-end; local alignment (BLAST, Smith-Waterman) finds the best-matching regions within longer sequences; multiple sequence alignment (ClustalW, MUSCLE) aligns three or more sequences simultaneously, often to study evolutionary relationships.

Genome Assembly

Reconstructing a complete genome from millions of short, overlapping DNA fragments produced by sequencing. De novo assembly (Velvet, SPAdes) builds a genome from scratch using only overlaps between fragments; reference-based assembly (BWA-MEM, Bowtie2) instead maps fragments onto an existing, related genome — much faster, but only works if a good reference already exists.

Statistical and Differential Expression Analysis

Once data is aligned or quantified, statistical methods determine whether observed differences are real or due to chance: hypothesis tests (t-tests, ANOVA), correlation analysis, and specialized tools like DESeq2 or edgeR that account for the specific noise characteristics of RNA-seq count data when identifying differentially expressed genes.

Network and Functional Analysis

Beyond individual genes, network analysis (using tools like Cytoscape or STRING) maps how genes, proteins, or metabolites interact as a system, while functional annotation (Gene Ontology, KEGG pathway mapping) assigns biological meaning to a list of genes that came out of a statistical test — turning "these 200 genes changed" into "this stress-response pathway was activated."

Why It Matters

A statistically significant gene list is not, by itself, a biological finding. It's the functional annotation and network analysis step that connects the numbers back to a coherent biological story a reader can actually evaluate.

Common Misunderstanding

Students often think a low p-value alone proves a result is biologically important. A tiny, biologically meaningless expression change can still produce a low p-value if the sample size is large enough — statistical significance and biological significance are related but distinct ideas, which is why effect size (e.g., fold-change) is reported alongside the p-value.

Key Terms

Term	Definition	Related Concept
High-Throughput Sequencing (NGS)	Technology generating massive volumes of DNA/RNA sequence data rapidly	scRNA-seq, ChIP-seq
Microarray	A chip with predefined probes used to measure gene expression or genetic variation	NGS
Mass Spectrometry	A technique identifying molecules by measuring their mass, used for proteins and metabolites	Proteomics
Flow Cytometry	A technique measuring properties of individual cells as they pass through a laser	FACS
Sequence Alignment	Comparing DNA/protein sequences to identify regions of similarity	BLAST, ClustalW
Genome Assembly	Reconstructing a full genome from overlapping sequence fragments	De novo Assembly, Reference-based Assembly
Differential Expression Analysis	Statistical comparison of gene expression levels between conditions	DESeq2, Fold-Change
Functional Annotation	Assigning known biological roles/pathways to a list of genes or proteins	Gene Ontology, KEGG

Common Mistakes

Misconception: More data collection technology (e.g., NGS instead of microarrays) automatically produces better, more trustworthy results. Why it's wrong: NGS produces more data and higher resolution, but it doesn't fix a flawed experimental design — a poorly controlled NGS experiment is still unreliable, just at higher cost and complexity. Correct understanding: The value of any collection technology depends entirely on the experimental design underneath it; better technology cannot substitute for proper controls and sample size.

---

Misconception: A statistically significant p-value (e.g., p < 0.05) on its own proves a gene is biologically important. Why it's wrong: Statistical significance only indicates the observed difference is unlikely to be due to random chance at that sample size — it says nothing about whether the magnitude of the change (effect size) is biologically meaningful, and with large enough sample sizes, trivial differences can become "significant." Correct understanding: Meaningful results should be judged using both statistical significance (p-value, often corrected for multiple testing) and effect size (e.g., fold-change), then interpreted through functional/pathway context.

---

Misconception: Reference-based genome assembly always works and is preferable to de novo assembly. Why it's wrong: Reference-based assembly requires a closely related, high-quality reference genome to map against; for a newly discovered or highly divergent organism, no adequate reference may exist, making the approach impossible or misleading. Correct understanding: De novo assembly is necessary when no suitable reference genome exists, while reference-based assembly is faster and more accurate when a good reference is available — the choice depends on what's available, not a fixed preference.

Comparison and Connections

Technology	Data Type	Resolution	Typical Cost	Best For
Microarray	Gene expression, SNPs	Lower (limited to known probes)	Lower	Targeted, well-characterized studies
NGS (RNA-seq/WGS)	Genomic, transcriptomic	High (unbiased, whole transcriptome/genome)	Higher	Discovery of novel variants/transcripts
Mass Spectrometry	Proteomic, metabolomic	High for molecule identification	Moderate-High	Protein/metabolite identification and quantification
Flow Cytometry	Single-cell properties	High per-cell resolution	Moderate	Cell population characterization/sorting

Assembly Approach	Requires Reference?	Speed	Best Use Case
De novo Assembly	No	Slower, computationally heavy	Novel or highly divergent organisms
Reference-based Assembly	Yes	Faster	Organisms closely related to an existing reference genome

Practice Questions

Recall

1. Name four major data types generated in biotechnology research. Look for: any four of genomic, transcriptomic, proteomic, metabolomic, epigenetic, structural data.

2. What is the difference between global and local sequence alignment? Look for: global alignment compares sequences end-to-end (Needleman-Wunsch); local alignment finds the best-matching subregions within longer sequences (BLAST, Smith-Waterman).

Understanding

3. Explain why a microarray might miss an important genetic finding that RNA-seq would catch. Look for: microarrays only measure predefined probes chosen in advance, so any novel transcript, splice variant, or unexpected sequence not represented on the array will go undetected, whereas RNA-seq captures the whole transcriptome without needing to predefine targets.

4. Why is functional annotation (e.g., Gene Ontology) necessary after a differential expression analysis, rather than stopping at the gene list? Look for: a raw list of statistically significant genes doesn't explain what biological process is happening; functional annotation groups genes into known pathways/functions, turning a list of names into an interpretable biological story.

Application

5. A researcher has a newly discovered bacterium with no closely related reference genome available. Which genome assembly approach should they use, and why? Look for: de novo assembly, because no suitable reference genome exists to map reads against.

6. A study finds 500 genes with p < 0.05 between two conditions but doesn't report fold-change. What additional information should a reviewer request, and why? Look for: fold-change (effect size) for each gene, because p-value alone doesn't indicate whether the expression differences are large enough to be biologically meaningful, especially with large sample sizes.

Analysis

7. Compare using flow cytometry versus mass spectrometry for studying a mixed population of immune cells producing different proteins. Which would you choose for which sub-question? Look for: flow cytometry is better for characterizing/sorting cell populations by surface markers and counting cell types; mass spectrometry is better for identifying and quantifying the specific proteins those cells produce — often both are used together for a complete picture.

8. A published RNA-seq study did not validate its top differentially expressed genes with an independent method like qRT-PCR. Evaluate the strength of its conclusions. Look for: the conclusions are weaker without independent validation, because RNA-seq and its analysis pipeline (alignment, normalization, statistical model) can produce false positives due to technical artifacts; qRT-PCR validation on a subset of genes increases confidence that the findings reflect true biology rather than pipeline noise.

FAQ

Q: Why do researchers still use microarrays when NGS is more powerful? Cost and simplicity. For well-studied organisms where the relevant genes are already known, a targeted microarray is cheaper and faster to analyze than a full RNA-seq experiment, and it avoids the heavier computational burden of NGS data processing.

Q: What's the practical difference between genomic and transcriptomic data? Genomic data describes the fixed DNA sequence (the "blueprint"), which is largely the same across all cells of an organism. Transcriptomic data describes which genes are actively being transcribed into mRNA at a given moment in a given cell type or condition — it changes based on context, while the genome mostly doesn't.

Q: Do I need to understand statistics to do bioinformatics data analysis? Yes, at least the fundamentals. Nearly every analysis tool (DESeq2, GWAS software, clustering algorithms) is built on statistical principles like hypothesis testing and multiple-testing correction; using these tools correctly requires understanding what a p-value and effect size actually mean.

Q: Why is single-cell RNA-seq (scRNA-seq) considered a major advance over standard "bulk" RNA-seq? Bulk RNA-seq averages gene expression across every cell in a sample, which can hide important differences between cell subtypes. scRNA-seq measures expression in individual cells, revealing heterogeneous cell populations that bulk sequencing would blend together into a misleading average.

Q: What should I do if my analysis pipeline gives a "significant" result but it seems biologically implausible? Don't report it as-is. Check for common technical artifacts first — batch effects, contamination, mismapped reads, or an inappropriate statistical test — and consider independent validation (e.g., qRT-PCR) before trusting a surprising result.

Quick Revision

• Data types in biotechnology include genomic, transcriptomic, proteomic, metabolomic, epigenetic, and structural data — each needs matched collection technology.
• Experimental design (controls, randomization) determines data quality before any collection technology is chosen; better technology cannot fix a flawed design.
• NGS gives high-resolution, unbiased sequence data; microarrays are cheaper but limited to predefined probes.
• Mass spectrometry (LC-MS, GC-MS, MS/MS) is the primary tool for identifying and quantifying proteins and metabolites.
• Flow cytometry (and FACS) measures and sorts individual cells by size, granularity, and fluorescence.
• Sequence alignment (global: Needleman-Wunsch; local: BLAST) is the foundation for comparing genomic/protein sequences.
• Genome assembly is de novo (no reference needed, slower) or reference-based (needs a related genome, faster).
• Statistical significance (p-value) and biological significance (effect size/fold-change) are different — both matter.
• Functional annotation and network analysis turn a raw gene list into an interpretable biological story.
• Independent validation (e.g., qRT-PCR after RNA-seq) strengthens confidence that a finding isn't a pipeline artifact.

Related Topics

Prerequisites: Introduction to Research Methodology, Research Design and Planning

Related Topics: Statistical Tools for Research, Writing Research Papers

Next Topics: Statistical Tools for Research, Research Ethics