# How to Present Your Bioinformatics Project Results By the end of this guide, you will be able to organize a bioinformatics project into a clean, shareable directory structure, generate a quality-control summary report with MultiQC, produce publication-ready figures in R, write a results narrative modeled on published bioinformatics papers, assemble a slide deck for a lab meeting or thesis committee, and package your code and computing environment so a colleague can rerun the entire analysis. You will need a Unix-like terminal (Linux, macOS, or Windows Subsystem for Linux), Python 3.8 or later with MultiQC installed, R 4.0 or later with the ggplot2 package, and Git for version control. Every command shown here runs on a standard laptop without specialized hardware. ## What Presenting Bioinformatics Results Involves Presenting bioinformatics results means more than pasting plots into PowerPoint. A bioinformatics project typically produces dozens of intermediate files, QC metrics from multiple tools, statistical tables, and a handful of final figures. Your job is to distill that into deliverables a specific audience can act on. Those deliverables fall into four categories. First, a written report. This is a structured document (HTML, PDF, or Word) that walks the reader through what you did, what you found, and what it means. Published bioinformatics papers such as the ulcerative colitis immune-infiltration study by Xue and colleagues [1] or the pulmonary fibrosis ferroptosis analysis by He and colleagues [2] follow a predictable arc: data source, [preprocessing](/blog/data-analysis/what-is-data-preprocessing), differentially expressed gene identification, functional enrichment, network analysis, and validation. Your report should mirror that arc even if your project is smaller in scope. Second, a slide deck. This is a condensed visual version of the report for a live audience. A lab-meeting deck might be 15 slides for a 20-minute talk. A thesis-committee deck might be 30 slides for a 45-minute presentation. The content overlaps with the report but the framing differs. Slides emphasize one message per slide. Reports allow nuance and caveats. Third, figures. These are the individual plots, heatmaps, network diagrams, and tables that carry your evidence. A well-designed figure answers a question without the reader needing to consult the text. Tools like XCluSim demonstrate how interactive visual analytics can help researchers compare multiple clustering results side by side, reducing the cognitive burden of assessing quality across algorithm outputs [3]. Fourth, a reproducibility package. This includes your scripts, a pinned software environment, a README explaining how to rerun the analysis, and version-controlled history. The reproducibility crisis in bioinformatics has driven initiatives like the Reprohackathon program at Université Paris-Saclay, which teaches Master's students to implement reproducible analyses over a three- to four-month project reanalyzing published data [4]. Tools such as QIIME 2 Provenance Replay go further by automatically generating executable code from prior results, removing much of the barrier to reproducing an analysis [5]. Each deliverable serves a different audience at a different level of detail. The rest of this guide walks through how to produce all four. ## How to Present Bioinformatics Results Step by Step ![Flow of seven steps: organize directory, run MultiQC, make figures, write narrative, build slides, package code, rehearse and deliver.](/article-images/illustrations/how-to-present-bioinformatics-results-seven-step-workflow.webp) This section gives a numbered walkthrough. Each step includes the commands to run, what the expected output looks like, and what to verify before moving on. ### Step 1: Organize Your Project Directory A disorganized project folder makes every downstream step harder. You will lose track of which figure came from which script, which version of the data you used, and whether that CSV in the root directory is a final result or a draft. Start by creating a standard structure. ```bash mkdir -p myproject/{data/raw,data/processed,results,figures,scripts,reports,docs} cd myproject ``` There is no output from this command. Verify the structure exists by listing the contents. ```bash ls -la ``` Expected output: ``` drwxr-xr-x 3 user group 96 Jan 15 10:00 data drwxr-xr-x 2 user group 64 Jan 15 10:00 docs drwxr-xr-x 2 user group 64 Jan 15 10:00 figures drwxr-xr-x 2 user group 64 Jan 15 10:00 reports drwxr-xr-x 2 user group 64 Jan 15 10:00 results drwxr-xr-x 2 user group 64 Jan 15 10:00 scripts ``` Place raw sequencing files or downloaded datasets in `data/raw`. Store cleaned, filtered, or normalized data in `data/processed`. Save numerical outputs (gene lists, differential expression tables, enrichment results) in `results`. Save rendered plots in `figures`. Save your R or Python scripts in `scripts`. Save rendered reports (HTML, PDF) in `reports`. Use `docs` for your README, a data dictionary, and any protocol notes. The reason this matters for presentation is simple. When your collaborator asks "which figure shows the enrichment results?", you can point to `figures/enrichment_dotplot.pdf` without searching. When a reviewer asks for the exact code that produced a number, you open `scripts/03_differential_expression.R`. ### Step 2: Generate a Quality-Control Summary Report Before you present any biological finding, you need to show that your input data passed QC. MultiQC aggregates output from many bioinformatics tools into a single interactive HTML report, enabling global trends and outlier samples to be identified quickly across an entire project [6]. If you ran FastQC on 24 RNA-seq samples, MultiQC combines all 24 reports into one page with overlaid per-base quality plots, duplication rates, and adapter contamination. Run MultiQC from your project root, pointing it at the directory containing your tool outputs. ```bash multiqc data/raw/fastqc_results -o reports/ ``` Expected output in the terminal: ``` /// MultiQC v1.21 | multiqc.info searching /path/to/myproject/data/raw/fastqc_results ... Found 24 FastQC reports Report written to: /path/to/myproject/reports/multiqc_report.html ``` Open `reports/multiqc_report.html` in a web browser. You will see a navigation sidebar listing each QC module (Sequence Quality, Per Base Sequence Content, Adapter Content, Overrepresented Sequences) and a main panel with interactive plots. Hover over any sample to see its values. Click the gear icon to customize thresholds. What to check before moving on: - Every sample appears in the report. If you expected 24 samples and see 23, one FastQC run failed silently. - Per-base quality scores stay above Q30 for the majority of reads. A sharp drop in the last few bases is normal for many sequencing platforms. - Adapter contamination is low. If one sample shows 40 percent adapter content while others show 2 percent, flag it as an outlier and investigate before proceeding. If MultiQC reports "No analysis results found," the directory path is wrong or the files do not match any recognized tool format. Double-check that you are pointing at the directory containing the actual FastQC output files (files ending in `_fastqc.zip` or `_fastqc.html`), not the parent directory. ### Step 3: Create Publication-Quality Figures Figures are the core of your presentation. A volcano plot showing differentially expressed genes is the single most recognizable figure in [transcriptomics](/knowledge/bioinformatics/modern-transcriptomics-bulk-single-cell-spatial) papers. Papers such as the breast cancer RNA-binding protein study by Wang and colleagues identified 90 upregulated and 115 downregulated RBPs and visualized them with volcano plots before proceeding to enrichment and network analysis [7]. Here is how to produce one in R with ggplot2. First, load your differential expression results. Assume you have a data frame called `de_results` with columns `gene`, `log2FoldChange`, and `padj` (adjusted p-value). ```r library(ggplot2) # Add a column for significance labeling de_results$threshold <- ifelse(de_results$padj < 0.05 & abs(de_results$log2FoldChange) > 1, "Significant", "Not significant") # Create the volcano plot p <- ggplot(de_results, aes(x = log2FoldChange, y = -log10(padj), color = threshold)) + geom_point(size = 1.5, alpha = 0.6) + scale_color_manual(values = c("Significant" = "#E74C3C", "Not significant" = "grey60")) + geom_hline(yintercept = -log10(0.05), linetype = "dashed", color = "black") + geom_vline(xintercept = c(-1, 1), linetype = "dashed", color = "black") + labs(x = "log2 Fold Change", y = "-log10 Adjusted P-value", color = "Threshold") + theme_minimal() + theme(legend.position = "bottom") # Save to file ggsave("figures/volcano_plot.pdf", plot = p, width = 6, height = 5, dpi = 300) ``` This produces a PDF file at `figures/volcano_plot.pdf`. Open it and verify the following: - Red points sit in the upper-left and upper-right quadrants, representing genes with large fold change and statistical significance. - Grey points cluster near the center, representing genes with minimal change. - Dashed lines mark the thresholds you chose (padj less than 0.05 horizontally, absolute log2 fold change greater than 1 vertically). A common error with ggplot2 is saving a blank or empty plot. This happens when you call `ggsave()` after the plot has already been rendered to the screen with `print(p)` or by typing `p` at the console. The fix is to store the plot object in a variable (`p <- ggplot(...)`) and pass that variable to `ggsave()`. Never call `ggsave()` after displaying the plot interactively. Another frequent issue is overlapping text labels when annotating gene names. The `ggrepel` package solves this by repelling labels away from each other. ```r install.packages("ggrepel") library(ggrepel) # Label the top 10 most significant genes top_genes <- head(de_results[order(de_results$padj), ], 10) p_labeled <- p + geom_text_repel(data = top_genes, aes(label = gene), size = 3, segment.color = "grey40") ggsave("figures/volcano_labeled.pdf", plot = p_labeled, width = 7, height = 5, dpi = 300) ``` ### Step 4: Write the Results Narrative Your results section tells the story of what you found in the order the reader needs to understand it. Study how published bioinformatics papers structure this. The ulcerative colitis study by Xue and colleagues screened 248 differentially expressed genes, narrowed to 20 immune-related hub genes, identified four candidate diagnostic biomarkers (DPP10, S100P, AMPD1, and ASS1), and then reported immune cell infiltration differences [1]. Each result builds on the previous one. The reader is never asked to accept a conclusion without seeing the evidence that led there. Follow this paragraph-level structure in your own write-up. Paragraph 1: Data description and QC summary. State the data source, sample sizes, and any filtering steps. Reference the MultiQC report. For example: "RNA-seq data from 24 tumor and 24 matched normal samples were obtained from GEO accession GSE12345. All samples passed QC with per-base quality scores above Q30 and adapter contamination below 5 percent." Paragraph 2: Primary result. State the main finding with the number. For example: "Differential expression analysis identified 156 upregulated and 203 downregulated genes at an adjusted p-value threshold of 0.05 and absolute log2 fold change greater than 1." Paragraph 3: Functional characterization. Describe enrichment results. For example: "Gene Ontology enrichment of the upregulated gene set revealed significant terms related to immune response and extracellular matrix organization." Paragraph 4: Validation or secondary analysis. If you validated findings with qRT-PCR, an independent dataset, or a survival analysis, report it here. Paragraph 5: Interpretation. Connect the results back to the biological question. Keep this brief in the results section. Deeper interpretation belongs in the Discussion. Write in past tense. Use active voice where possible. Avoid vague qualifiers like "somewhat" or "relatively." If a fold change is 3.2, write 3.2. ### Step 5: Assemble the Slide Deck A slide deck is not the report cut into pieces. It is a separate artifact designed for a live audience with limited attention. Follow these rules. One message per slide. If your slide title is "Differential expression reveals 156 upregulated genes," the slide should show the volcano plot and nothing else. Move the full gene list to a supplementary table. Show the pipeline early. Audiences unfamiliar with your project need a slide that shows the overall workflow: data acquisition, preprocessing, analysis, validation. This slide orients them for everything that follows. Lead with the answer. State the conclusion first, then show the supporting evidence. Do not build suspense. Scientists in your audience want to know where you are going so they can evaluate each piece of evidence in context. Use consistent figure formatting. Every figure should use the same font size, color palette, and axis label style. Inconsistent formatting signals carelessness and distracts the audience. Limit text. If a slide has more than 30 words of body text, the audience will read instead of listen. Move detailed text to speaker notes. ### Step 6: Package Code and Dependencies for Reproducibility A presentation without a reproducibility package is a conversation, not a scientific contribution. The ISCB competency framework identifies reproducibility as a core skill for bioinformatics professionals [8]. Two actions make your project reproducible: pinning the software environment and version-controlling the code. Export your conda environment so anyone can recreate the exact software versions you used. ```bash conda env export > environment.yml ``` Expected output is a YAML file. The first few lines look like this: ```yaml name: bioinfo_env channels: - conda-forge - bioconda dependencies: - python=3.11.5 - multiqc=1.21 - r-base=4.3.1 - r-ggplot2=3.4.4 ``` Initialize a Git repository and commit your work. ```bash git init git add scripts/ environment.yml README.md git commit -m "Initial analysis: QC, differential expression, enrichment" ``` Verify the commit history. ```bash git log --oneline ``` Expected output: ``` a1b2c3d Initial analysis: QC, differential expression, enrichment ``` Write a `README.md` file in the `docs` directory explaining how to rerun the analysis from scratch. Include the conda installation command, the order in which scripts should be run, and the expected runtime. The Reprohackathon project found that implementing reproducible analyses requires significant effort but yields substantial benefits for both the analyst and the broader community [4]. Automated provenance-tracking tools like QIIME 2 Provenance Replay reduce this burden by recording analysis steps as metadata and regenerating executable code from stored results [5]. ### Step 7: Rehearse and Deliver Practice your presentation out loud at least twice. Time yourself. If a 20-minute slot turns into 35 minutes of rehearsal, cut slides or tighten explanations. Identify the three questions you are most likely to face and prepare answers. Those questions usually target sample size, statistical methodology, and biological interpretation. During the actual presentation, point at the figure when explaining it. Do not read the slide. If someone asks a question you cannot answer, say "I do not have that number handy, but I can follow up after the talk." Write the question down. Follow up. ## Matching Figure Types to Result Types Choosing the wrong figure type is one of the most common ways bioinformatics presentations fail. The table below maps common result types to recommended visualizations. | Result type | Recommended figure | Key elements to label | Common tool | |--|--|--|--| | Differential expression | Volcano plot | Gene names for top hits, fold-change and p-value thresholds | DESeq2, ggplot2 | | Enrichment analysis | Dot plot or bar chart | Pathway name, gene ratio, adjusted p-value | clusterProfiler | | Sample clustering | Heatmap or PCA plot | Sample IDs, cluster assignments, color scale legend | pheatmap, ComplexHeatmap | | QC metrics | Multi-panel summary | Per-sample distributions, failure thresholds | MultiQC, FastQC | | Protein interaction network | Node-edge diagram | Hub genes, edge confidence, community clusters | Cytoscape, STRING | | Survival association | Kaplan-Meier curve | Risk groups, p-value, number at risk | survival, survminer | When your analysis produces multiple clustering results that need side-by-side comparison, tools like XCluSim provide interactive visual analytics designed specifically for comparing multiple clustering ensembles, supporting the Visual Information Seeking Mantra of overview first, zoom and filter, then details on demand [3]. For most presentations, a static heatmap suffices. Reserve interactive tools for exploratory analysis rather than final presentation. Web-based platforms such as the EMBL-EBI sequence analysis toolkit provide built-in result visualizations for sequence similarity searches and multiple sequence alignments through a consistent interface, which reduces the formatting burden when presenting results from those services [9]. ## Adapting Your Presentation to the Audience ![Decision path asking who the audience is, leading to lab meeting, clinical collaborator or thesis committee framing.](/article-images/illustrations/how-to-present-bioinformatics-results-audience-framing.webp) The same results require different framing depending on who is listening. A lab-meeting audience of computational biologists wants methodological detail. A clinical collaborator wants the biological implication. A funding committee wants the impact. For a lab meeting, spend 30 percent of your time on methods. Show your pipeline. Explain parameter choices. Discuss what you tried that did not work. This audience will probe your methodology, so preempt their questions. For a clinical or domain-science audience, minimize methodological detail. Lead with the biological finding. Use analogies. Instead of saying "we applied DESeq2 with a negative binomial model," say "we identified genes that behave differently between diseased and healthy tissue, accounting for natural variation between samples." For a thesis committee or external seminar, balance both. Show that you understand the methods deeply enough to defend them but focus most of your time on the biological insight. The ISCB competency framework version 3 defines the knowledge, skills, and attitudes expected at different career stages in bioinformatics, including the ability to communicate results to diverse audiences [8]. If you are building your skills through a [bioinformatics course](/blog/careers/bioinformatics-courses-and-certifications-a-practical-guide) or self-directed learning, practice presenting the same results to two different audiences and notice what changes. ## Writing Style for Bioinformatics Results Bioinformatics writing has conventions that differ from general scientific prose. Numbers go in parentheses after the claim, not buried in the sentence. Statistical thresholds are stated explicitly. Tool versions and parameters are recorded in the Methods section, not repeated in Results. Use precise language. "Enriched" means overrepresented beyond chance at a specified threshold. "Correlated" means a statistical association was measured. "Associated with" means a relationship was observed without implying mechanism. Do not write "these genes are involved in" unless you have functional evidence. Write "these genes are enriched for annotations related to" instead. Avoid anthropomorphizing genes. Genes do not "try" or "want" or "respond in order to." Genes show differential expression. Pathways show enrichment. Proteins show altered abundance. Keep sentences short. If a sentence has more than two clauses, split it. Your reader is processing complex information. Long sentences add cognitive load without adding precision. ## Common Mistakes and Limitations These are the errors that most frequently undermine bioinformatics presentations, drawn from reviewing practice and published critiques. Presenting raw tool output as a result. A table of 248 differentially expressed genes is not a result. It is an intermediate product. The result is what those genes tell you biologically. Always interpret. Say what the genes do, what pathways they enrich, and what that means for your hypothesis. Showing QC only if asked. Some analysts treat QC as internal housekeeping and omit it from presentations. Reviewers and collaborators need to see QC to trust everything downstream. Include a QC summary slide early in every presentation, even if it feels routine. Overloading slides with figures. Four heatmaps on one slide is unreadable. One heatmap per slide with a clear title stating the takeaway is readable. If you have multiple related figures, use a build (progressive disclosure) or move extras to backup slides. Ignoring batch effects and confounders. If your disease group and control group were sequenced in different batches, that confound must be addressed before you present differential expression results. Papers that skip this step risk attributing technical variation to biology. Always show a PCA or multidimensional scaling plot colored by batch before colored by condition. Using default color schemes that are not colorblind-safe. The default red-green color scheme in many tools is invisible to approximately 8 percent of men with red-green color vision deficiency. Use palettes from ColorBrewer or the viridis family, which are designed for accessibility. Not stating software versions. A result produced with DESeq2 version 1.38 may differ from one produced with version 1.42. Record versions in your Methods section and environment file. Presenting unvalidated computational predictions as confirmed findings. Many bioinformatics papers identify candidate biomarkers computationally and then validate a subset experimentally. The ulcerative colitis study by Xue and colleagues identified four candidate biomarkers through bioinformatics and validated them before drawing conclusions [1]. The pulmonary fibrosis study by He and colleagues used qRT-PCR in a mouse model to verify hub [gene expression](/blog/guides/gene-expression) identified through GEO data mining and [random forest](/blog/data-analysis/what-is-random-forest) screening [2]. If your project is purely computational, state that limitation clearly. Presenting results from a single dataset without replication. Findings from one cohort may not generalize. If an independent dataset is available, use it for validation. If not, acknowledge the limitation. Confusing [statistical significance](/blog/guides/statistical-significance) with biological significance. A gene with an adjusted p-value of 0.001 and a log2 fold change of 0.3 is statistically significant but may be biologically irrelevant. Present effect sizes alongside p-values. Forgetting to acknowledge that individual biological systems require domain-expert interpretation. A computational analysis identifies patterns. Translating those patterns into biological or clinical action requires collaboration with domain scientists who understand the system under study. ## Frequently Asked Questions ### How long should a bioinformatics results presentation be? A lab-meeting presentation should run 15 to 20 minutes with 10 to 15 slides. A conference talk is typically 10 to 15 minutes. A thesis committee update is 30 to 45 minutes. Match the number of slides to the time slot at roughly one slide per minute, minus a few slides for transitions. ### What software do I need to create a bioinformatics report? At minimum, you need a terminal, Python with MultiQC for QC summaries, R with ggplot2 for figures, and a text editor or Quarto for rendering reports. Git and conda are strongly recommended for reproducibility. All of these are free and open source. ### How many figures should I include in a bioinformatics presentation? Include one figure per main finding. A typical project has three to five main findings, so three to five core figures plus one QC summary and one methods-overview graphic. Supplementary figures go in backup slides or an appendix. ### How do I present preliminary or unfinished bioinformatics results? Label them clearly as preliminary on the title slide and on each relevant slide. State what is complete and what remains. Audiences respond well to honesty about limitations. Frame the presentation around the question you are addressing rather than the final answer you have not yet reached. ### What is the difference between a results section and a methods section? The methods section describes what you did and how, in enough detail that someone could reproduce it. The results section describes what you found, with no procedural detail. If you find yourself writing "we then ran" in the results section, move that sentence to methods. ### How do I make my bioinformatics analysis reproducible for others? Pin your software environment with conda or Docker, version-control your scripts with Git, write a README with step-by-step instructions, and store raw data or provide a download link. The Reprohackathon curriculum emphasizes that reproducibility requires deliberate effort across versioning, containerization, and [workflow management](/knowledge/bioinformatics/snakemake-vs-nextflow-workflow-architectures) [4]. ### Should I show code during my bioinformatics presentation? Rarely. Show code only when the coding decision itself is the point, such as explaining a custom statistical test or a nonstandard parameter choice. For most audiences, showing the pipeline diagram and the resulting figure communicates more effectively than showing the script. ### How do I handle questions about my bioinformatics results that I cannot answer? Say honestly that you do not have that information at hand, write the question down, and commit to following up. Speculating an answer damages credibility more than admitting uncertainty. If the question targets a methodological choice you did not make, explain that the choice was inherited from a protocol or collaborator and offer to investigate. ## Sources 1. [Characteristics of immune cell infiltration and associated diagnostic biomarkers in ulcerative colitis: results from bioinformatics analysis](https://pubmed.ncbi.nlm.nih.gov/33323040/) 2. [Bioinformatics Analysis Identifies Potential Ferroptosis Key Genes in the Pathogenesis of Pulmonary Fibrosis](https://pubmed.ncbi.nlm.nih.gov/35069688/) 3. [XCluSim: a visual analytics tool for interactively comparing multiple clustering results of bioinformatics data](https://pubmed.ncbi.nlm.nih.gov/26328893/) 4. [Reprohackathons: promoting reproducibility in bioinformatics through training](https://pubmed.ncbi.nlm.nih.gov/37387150/) 5. [Facilitating bioinformatics reproducibility with QIIME 2 Provenance Replay](https://pubmed.ncbi.nlm.nih.gov/38011287/) 6. [MultiQC: summarize analysis results for multiple tools and samples in a single report](https://pubmed.ncbi.nlm.nih.gov/27312411/) 7. [Integrated Bioinformatics Analysis the Function of RNA Binding Proteins (RBPs) and Their Prognostic Value in Breast Cancer](https://pubmed.ncbi.nlm.nih.gov/30881302/) 8. [The ISCB competency framework v. 3: a revised and extended standard for bioinformatics education and training](https://pubmed.ncbi.nlm.nih.gov/39678208/) 9. [A new bioinformatics analysis tools framework at EMBL-EBI](https://pubmed.ncbi.nlm.nih.gov/20439314/)