scientific-skills/pptx-posters/references/poster_content_guide.md
Content is king in research posters. This guide covers writing strategies, section-specific guidance, visual-text balance, and best practices for communicating research effectively in poster format.
Reality: Most viewers spend 3-5 minutes at your poster
Design Implication: Poster must work at three levels:
Poster ≠ Condensed Paper
Paper approach (❌):
Poster approach (✅):
Story Arc for Posters:
Hook (Problem) → Approach → Discovery → Impact
Example:
Word Count Guidelines:
Word Budget by Section:
| Section | Word Count | % of Total |
|---|---|---|
| Introduction/Background | 50-100 | 15% |
| Methods | 100-150 | 25% |
| Results (text) | 100-200 | 25% |
| Discussion/Conclusions | 100-150 | 25% |
| References/Acknowledgments | 50-100 | 10% |
Counting Tool:
% Add word count to poster (remove for final)
\usepackage{texcount}
% Compile with: texcount -inc poster.tex
Optimal Balance: 40-50% visual content, 50-60% text+white space
Visual Content Includes:
Too Text-Heavy (❌):
Well-Balanced (✅):
Purpose: Capture attention, convey topic, establish credibility
Characteristics of Effective Titles:
Title Formulas:
1. Descriptive:
[Method/Approach] for [Problem/Application]
Example: "Deep Learning for Early Detection of Alzheimer's Disease"
2. Question:
[Research Question]?
Example: "Can Microbiome Diversity Predict Treatment Response?"
3. Assertion:
[Finding] in [Context]
Example: "Novel Mechanism Identified in Drug Resistance Pathways"
4. Colon Format:
[Topic]: [Specific Approach/Finding]
Example: "Urban Heat Islands: A Machine Learning Framework for Mitigation"
Avoid:
LaTeX Title Formatting:
% Emphasize key words with bold
\title{Deep Learning for \textbf{Early Detection} of Alzheimer's Disease}
% Two-line titles for long names
\title{Machine Learning Framework for\\Urban Heat Island Mitigation}
% Avoid ALL CAPS (harder to read)
Best Practices:
Format Examples:
% Simple format
\author{\textbf{Jane Smith}\textsuperscript{1}, John Doe\textsuperscript{2}}
\institute{
\textsuperscript{1}University of Example,
\textsuperscript{2}Research Institute
}
% With contact
\author{Jane Smith\textsuperscript{1,*}}
\institute{
\textsuperscript{1}Department, University\\
\textsuperscript{*}[email protected]
}
Purpose: Establish context, motivate research, state objective
Structure (50-100 words):
Example (95 words):
Antibiotic resistance causes 700,000 deaths annually, projected to reach
10 million by 2050. Current diagnostic methods require 48-72 hours,
delaying appropriate treatment. Machine learning offers potential for
rapid resistance prediction, but existing models lack generalizability
across bacterial species.
We developed a transformer-based deep learning model to predict antibiotic
resistance from genomic sequences across multiple pathogen species. Our
approach integrates evolutionary information and protein structure to
improve cross-species accuracy.
Visual Support:
Common Mistakes:
Purpose: Describe approach sufficiently for understanding (not replication)
Key Question: "How did you do it?" not "How could someone else replicate it?"
Content Strategy:
Visual Methods (Highly Recommended):
% Flowchart of study design
\begin{tikzpicture}[node distance=2cm]
\node (start) [box] {Data Collection\\n=1,000 samples};
\node (process) [box, below of=start] {Preprocessing\\Quality Control};
\node (analysis) [box, below of=process] {Statistical Analysis\\Mixed Models};
\node (end) [box, below of=analysis] {Validation\\Independent Cohort};
\draw [arrow] (start) -- (process);
\draw [arrow] (process) -- (analysis);
\draw [arrow] (analysis) -- (end);
\end{tikzpicture}
Text Methods (50-150 words):
For Experimental Studies:
Methods
• Study design: Randomized controlled trial (n=200)
• Participants: Adults aged 18-65 with Type 2 diabetes
• Intervention: 12-week exercise program vs. standard care
• Outcomes: HbA1c (primary), insulin sensitivity (secondary)
• Analysis: Linear mixed models, intention-to-treat
For Computational Studies:
Methods
• Dataset: 10,000 labeled images from ImageNet
• Architecture: ResNet-50 with custom attention mechanism
• Training: 100 epochs, Adam optimizer, learning rate 0.001
• Validation: 5-fold cross-validation
• Comparison: Baseline CNN, VGG-16, Inception-v3
Format Options:
Purpose: Present key findings visually and clearly
Golden Rule: Show, don't tell
Content Allocation:
How Many Results:
Figure Selection Criteria:
Figure Captions:
Example Caption:
\caption{Treatment significantly improved outcomes.
Mean±SD shown for control (blue, n=45) and treatment (orange, n=47) groups.
**p<0.01, ***p<0.001 (two-tailed t-test).}
Text Support for Results (100-200 words):
Example Results Text:
Key Findings
• Model achieved 87% accuracy on test set (vs. 73% baseline)
• Performance consistent across 5 bacterial species (p<0.001)
• Prediction speed: <30 seconds per isolate
• Feature importance: protein structure (42%), sequence (35%),
evolutionary conservation (23%)
Data Presentation Formats:
1. Bar Charts: Comparing categories
\begin{tikzpicture}
\begin{axis}[
ybar,
ylabel=Accuracy (\%),
symbolic x coords={Baseline, Model A, Our Method},
xtick=data,
nodes near coords
]
\addplot coordinates {(Baseline,73) (Model A,81) (Our Method,87)};
\end{axis}
\end{tikzpicture}
2. Line Graphs: Trends over time 3. Scatter Plots: Correlations 4. Heatmaps: Matrix data, clustering 5. Box Plots: Distributions, comparisons 6. ROC Curves: Classification performance
Purpose: Interpret findings, state implications, acknowledge limitations
Structure (100-150 words):
1. Main Conclusions (50-75 words):
Example:
Conclusions
• First cross-species model for antibiotic resistance prediction
achieving >85% accuracy
• Protein structure integration critical for generalizability
(improved accuracy by 14%)
• Prediction speed enables clinical decision support within
consultation timeframe
• Potential to reduce inappropriate antibiotic use by 20-30%
2. Limitations (25-50 words, optional but recommended):
Example:
Limitations
• Training data limited to 5 bacterial species
• Requires genomic sequencing (not widely available)
• Validation needed in prospective clinical trials
3. Future Directions (25-50 words, optional):
Example:
Next Steps
• Expand to 20+ additional species
• Develop point-of-care sequencing integration
• Launch multi-center clinical validation study (2025)
Avoid:
How Many: 5-10 key citations
Selection Criteria:
Format: Abbreviated, consistent style
Examples:
Numbered (Vancouver):
References
1. Smith et al. (2023). Nature. 615:234-240.
2. Jones & Lee (2024). Science. 383:112-118.
3. Chen et al. (2022). Cell. 185:456-470.
Author-Year (APA):
References
Smith, J. et al. (2023). Title. Nature, 615, 234-240.
Jones, A., & Lee, B. (2024). Title. Science, 383, 112-118.
Minimal (For Space Constraints):
Key References: Smith (Nature 2023), Jones (Science 2024),
Chen (Cell 2022). Full bibliography: [QR Code]
Alternative: QR code linking to full reference list
Include:
Format (25-50 words):
Acknowledgments
Funded by NIH Grant R01-123456 and NSF Award 7890123.
We thank Dr. X for data access, the Y Core Facility for
sequencing, and Z for helpful discussions.
Essential Elements:
Format:
Contact: Jane Smith, [email protected]
Lab: smithlab.university.edu | Twitter: @smithlab
QR Code Alternative:
Prefer Active Voice (more engaging, clearer):
Passive Voice (when appropriate):
Keep Sentences Short:
Example Revision:
Use Bullet Points For:
Use Short Paragraphs For:
Bullet Point Best Practices:
Example:
Methods
• Participants: 200 adults (18-65 years)
• Design: Double-blind RCT (12 weeks)
• Intervention: Daily 30-min exercise
• Control: Standard care
• Analysis: Mixed models (SPSS v.28)
First Use Rule: Define at first appearance
We used machine learning (ML) to analyze... Later, ML predicted...
Common Acronyms: May not need definition if universal to field
Avoid Excessive Jargon:
Present Statistics Clearly:
Format Numbers:
Example:
Treatment increased response by 23.5% (95% CI: 18.2-28.8%, p<0.001, n=150)
Figure First, Text Second:
Text Placement Relative to Figures:
On-Figure Annotations:
\begin{tikzpicture}
\node[inner sep=0] (img) {\includegraphics[width=10cm]{figure.pdf}};
\draw[->, thick, red] (8,5) -- (6,3) node[left] {Key region};
\draw[red, thick] (3,2) circle (1cm) node[above=1.2cm] {Anomaly};
\end{tikzpicture}
Callout Boxes:
\begin{tcolorbox}[colback=yellow!10, colframe=orange!80,
title=Key Finding]
Our method reduces errors by 34\% compared to state-of-the-art.
\end{tcolorbox}
Visual Section Markers:
\usepackage{fontawesome5}
\block{\faFlask~Introduction}{...}
\block{\faCog~Methods}{...}
\block{\faChartBar~Results}{...}
\block{\faLightbulb~Conclusions}{...}
Condensation Process:
1. Identify Core Message (The Elevator Pitch):
2. Select Key Results:
3. Simplify Methods:
4. Trim Literature Review:
5. Condense Discussion:
Specialist Audience (Same Field):
General Scientific Audience:
Public/Lay Audience:
Example Adaptation:
Specialist: "CRISPR-Cas9 knockout of BRCA1 induced synthetic lethality with PARP inhibitors"
General: "We used gene editing to make cancer cells vulnerable to existing drugs"
Public: "We found a way to make cancer treatments work better by targeting specific genetic weaknesses"
Clarity:
Completeness:
Accuracy:
Engagement:
Distance Test:
Scan Test:
Detail Test:
1. Too Much Text
2. Unclear Message
3. Methods Overkill
4. Poor Figure Integration
5. Missing Context
Effective poster content:
Remember: Your poster is a conversation starter, not a comprehensive treatise. Design content to intrigue, engage, and invite discussion.