Scientific Slide Design

This skill turns a research story into a deck that an audience can read in real time. The goal is comprehension at speaking pace: every slide communicates one idea in the time it takes to say one sentence. It complements ors-scientific-visualization-figure-design (data plots for the slides) and ors-scientific-visualization-schematics-diagrams (workflow / study-design cartoons).

When to use

Trigger this skill when any of the following apply:

• You are preparing a conference talk, departmental seminar, journal club, or thesis defense.
• You are pitching a grant or paper to a non-specialist audience.
• You are recording an asynchronous talk for a webinar, MOOC, or YouTube."
• A co-author or mentor has flagged a slide as "too busy" or "I couldn't read the bottom row".
• You are converting a paper into a 10–15 minute talk.

When NOT to use

• Posters — see ors-scientific-visualization-poster-design.
• Print figures for a paper — see ors-scientific-visualization-figure-design.
• Schematics, workflows, study design — see ors-scientific-visualization-schematics-diagrams.
• Slide-decks that are pure text (manuscript, README) — use ors-scientific-writing instead.

Prerequisites

• A defined audience and venue (specialist conference? departmental seminar? high-school outreach?). This decides depth, jargon, and visual style.
• A target talk length in minutes, and a rule of thumb: ~1 slide per minute for a fast conference talk, 1 slide per 2 minutes for a lecture.
• A list of 3–5 main messages. The whole deck exists to deliver these.
• The figures from the paper (PDF, PNG) or the data and code to regenerate them.
• A backup plan: a printed handout, a PDF copy on a USB stick, and a version that does not depend on network access at the venue.

Core workflow

1. Anchor on the one-sentence takeaway. Write the conclusion the audience should walk out remembering. The whole deck argues for that sentence.
2. Results-first structure. Use a problem → approach → key result → implication arc, not a chronological "we did X, then Y". Most slides should be results; the introduction and methods should be short.
3. Apply the 6×6 rule. No more than 6 lines per slide, no more than 6 words per line. If a slide is busier than this, split it.
4. Apply the 5-second test. Show a slide to a colleague for 5 seconds, then hide it. If they cannot state the main idea, simplify or rewrite the headline.
5. Use one main idea per slide. A slide is a single claim, supported by one figure, one table, or one bulleted list. Two claims = two slides.
6. Use the headline-as-conclusion pattern. Replace "Methods" with "Single-cell RNA-seq of 50,000 cells across 12 patients". The slide title is the message, not the topic.
7. Build progression for complex figures. Use 2–4 sequential reveals (one panel at a time, one arrow at a time). The audience reads at the pace you build, not all at once.
8. Restrain animation. Use animation for build progression only. No spinning logos, no fly-ins, no random slide transitions. Build once, then leave the final state on screen while you speak.
9. Cite on the slide, not just at the end. A small citation footer (Smith et al., Nature 2024) lets the audience follow up without breaking the talk.
10. Audit for the venue. Projector contrast, color profile, and aspect ratio differ by room. Test the deck on the actual projector or a similar one the day before.

Code patterns

A. Tool selection matrix

B. Beamer — minimal conference talk skeleton

\documentclass[10pt,aspectratio=169]{beamer}
\usetheme{metropolis}              % clean, modern, sans-serif
\usepackage[utf8]{inputenc}
\usepackage{graphicx}
\usepackage{booktabs}
\usepackage{hyperref}

\title{One-Sentence Takeaway}
\author{Pradyumna Jayaram}
\institute{Institution}
\date{Conference Name, 2026}

\begin{document}
\maketitle

\section{Introduction}
\begin{frame}{The problem}
    % Headline is the message, not the topic
    Single-cell assays are noisy; bulk averages hide the cells that matter.
    \begin{itemize}
        \item ~50{,}000 cells, 12 patients
        \item Goal: find a rare population driving outcome
    \end{itemize}
\end{frame}

\section{Approach}
\begin{frame}{Cohort and assay}
    \begin{columns}
        \begin{column}{0.5\textwidth}
            \begin{itemize}
                \item 10x Genomics 3' v3
                \item 12 patients, 4 conditions
                \item QC with `cellranger` and `scrublet`
            \end{itemize}
        \end{column}
        \begin{column}{0.5\textwidth}
            \includegraphics[width=\textwidth]{workflow.png}
        \end{column}
    \end{columns}
\end{frame}

\section{Results}
\begin{frame}{A rare population predicts outcome}
    \begin{figure}
        \centering
        \includegraphics[width=0.8\textwidth]{umap_rare.pdf}
        \caption{\textbf{Rare cluster (cyan) is 1.2\% of cells} but
                 explains 38\% of outcome variance (OR 4.7, 95\% CI 2.1--10.5).}
    \end{figure}
\end{frame}

\section{Conclusion}
\begin{frame}{Take-home}
    \begin{alertblock}{}
        A rare, transcriptionally distinct cell population
        drives patient outcome and is a candidate therapeutic target.
    \end{alertblock}
    \begin{itemize}
        \item Validation in an independent cohort is ongoing
        \item Code: \texttt{github.com/yourname/project}
    \end{itemize}
\end{frame}

\begin{frame}{References}
    \footnotesize
    \bibliographystyle{plain}
    \bibliography{refs}
\end{frame}
\end{document}

Compile with pdflatex talk.tex (twice for the table of contents) or latexmk -pdf talk.tex.

C. Marp — Markdown to slides (code-heavy talks)

---
marp: true
theme: default
paginate: true
size: 16:9
style: |
  section.lead h1 { text-align: center; }
  section { font-size: 26px; }
---

# Single-cell RNA-seq identifies a rare outcome-driving population
### Pradyumna Jayaram — Conference 2026

---

## The problem

Single-cell assays are noisy; bulk averages hide the cells that matter.

- 50,000 cells, 12 patients
- Goal: rare population driving outcome

---

## A rare population predicts outcome

![width:800px](umap_rare.png)

**Rare cluster (cyan, 1.2% of cells) explains 38% of outcome variance** (OR 4.7, 95% CI 2.1–10.5).

Render with marp talk.md --pdf or marp talk.md -o talk.html for a web-friendly version.

D. Quarto + reveal.js — reproducible report-as-talk

---
title: "Single-cell RNA-seq identifies a rare outcome-driving population"
author: "Pradyumna Jayaram"
format:
  revealjs:
    theme: simple
    incremental: true
    slide-number: true
    chalkboard: true
bibliography: refs.bib
jupyter: python3
---

# Code chunks render and execute; figures embed directly
import scanpy as sc
adata = sc.read_h5ad("data.h5ad")
sc.tl.umap(adata)
sc.pl.umap(adata, color="leiden", save="umap_leiden.png")

## Results

![](umap_leiden.png){fig-align="center" width="70%"}

A rare cluster (cyan) explains 38% of outcome variance [@smith2024singlecell].

Build with quarto render talk.qmd → talk.html. The same source can be re-rendered as a PDF, a manuscript, or a poster.

E. PowerPoint / Keynote — build progression (one panel at a time)

Pseudocode for a 4-panel reveal:

Slide "Approach"
  Build 1: panel A (cohort diagram)
  Build 2: + panel B (assay workflow)
  Build 3: + panel C (QC steps)
  Build 4: + panel D (statistical model)
  [leave the final state on screen while you speak]

In PowerPoint, set each panel to appear on click via the "Animations" pane; choose "Appear" or "Fade", not "Fly In" or "Spin". In Keynote, use "Build In" with "Fade".

F. Citation footer (consistent across the deck)

PowerPoint / Keynote: insert a small text box anchored to the bottom-left of the slide master, font 9–10 pt, italic. Content: Smith et al., Nature 2024 or \cite{smith2024} if generated by a Beamer bibliography.

Beamer: add to the footer template with \setbeamertemplate{footline}{...}.

Marp / Quarto: define a global footer in the frontmatter.

The 6×6 rule, in practice

The 6×6 rule (no more than 6 lines, 6 words per line) is a heuristic, not a law. The deeper rule is readability at glance:

• Title slide: 1–2 lines of title, 1 line of name, 1 line of affiliation.
• Content slide: 1 title (the message), ≤ 6 bullet lines, ≤ 6 words per line, supporting figure.
• Methods slide: short prose (1–2 sentences) + 1 figure, not both a wall of text and a figure.
• "Q&A" slide: 1 line ("Questions?") + an email and a QR code to the paper.

If a slide needs 8 bullets, split it into 2. If a bullet needs 14 words, cut it.

Results-first structure (a defensible default)

A common talk template for a 12-minute conference talk:

The exact slide count varies; the ratio does not. ~60% results, ~20% introduction + approach, ~20% conclusion + Q&A.

Animation: when to use it, when not

• Use it: to reveal a multi-panel figure one panel at a time; to underline a key element with an arrow or box; to show progression (year 1 → year 5 of a cohort).
• Avoid it: spinning logos, random slide transitions, fly-ins, typewriter text, animated charts, "click here!" pointers, auto-advancing slides, embedded video without a backup screenshot.
• The "leave it on screen" rule. After a build sequence, the final state stays on screen while you speak about it. The audience should not have to remember the order of reveals.

Dark mode for venues

Many conference rooms and lecture halls are dark. Choose slide background based on the venue:

• Light background, dark text is the safest default for mixed venues and printed handouts. Use a slight off-white (#FAFAFA) instead of pure white to reduce eye strain.
• Dark background, light text works for projector-only rooms where the lights are off. Use a near-black (#1A1A1A) and an off-white text (#EAEAEA) to avoid the glare of pure black/white.
• Test contrast on the actual projector. A 4.5:1 contrast ratio on a laptop may look fine but wash out on a projector with a yellow tint. Run the WCAG contrast check at projector gamma, not at laptop gamma.
• Avoid mid-tones for text. Mid-gray (#888888) text on a gray background is illegible under projector light. Use ≥ 7:1 contrast for body text and ≥ 4.5:1 for fine print.

Citations on the slide

There is no universal style, but the conventions in major venues:

• APA / Nature numeric: small superscript number on the slide, full reference in a final slide or handout ([1] Smith et al., Nature 2024).
• Author-year: inline Smith et al. (2024) in the body of the bullet, full reference list at the end.
• For data and code: include the accession or repository on the slide (GEO: GSE123456, github.com/yourname/repo).
• Avoid hyperlinks in the slide body. Hyperlinks look fine on a laptop and become unreadable on a projector. Put the URL in the speaker notes or the handout.

Accessibility for slides

• Font size. Body text ≥ 24 pt for the back row of a typical room; ≥ 28 pt for a large auditorium. Titles ≥ 32 pt.
• Color + redundancy. Encode groups by color + shape, never by color alone. Aim for ≥ 4.5:1 contrast (WCAG 2.1 AA) for body text.
• Alt text. Add alt text to every figure in the deck (PowerPoint: right-click → "Edit Alt Text"; Keynote: "Format → Advanced → Alternative Text"). The alt text should describe the figure in 1–2 sentences, not just its filename.
• Reading order. Use the slide outline, not visual stacking, as the reading order. Screen readers and pdftotext extract the outline first.
• No flashing content. Avoid strobing or rapid flashing; WCAG 2.1 caps flashing at 3 Hz.
• Captions on embedded video. If the talk has a video, caption it. Auto-captions are a starting point, not a final answer.
• Speaker notes for live captioning. Many venues run automated captioning. Speaker notes give the captioner context.

Common pitfalls

• The "manuscript on a slide" anti-pattern: copying a figure caption or a paragraph of the paper onto a slide. The slide is a medium, not a copy; rewrite for spoken pace.
• Title = topic, not message: a slide titled "Results" tells the audience nothing. Rename to "A rare population predicts outcome".
• Too many figures on one slide: 4 small figures, each illegible. Show one big figure with 2 builds.
• Body text 14 pt or smaller: invisible from the back row. The minimum is 24 pt; default to 28.
• Truncated axis without a break mark: a bar chart that starts at 30% exaggerates the difference. Mark the break.
• Inconsistent style across slides: one slide in Times New Roman, the next in Calibri, the next in Arial. Set a master / theme and use it.
• Embedded video that does not play in the venue: a codec the projector laptop lacks, a YouTube link behind a firewall. Always have a still-frame screenshot as a backup.
• Speaker notes that read the slide: notes are a script, not a copy. If the slide and the notes say the same thing, the slide is too busy.
• "Thank you / Questions?" without contact info: an audience member wants to follow up but has no email. Add the email and a QR code to the paper.

Validation

• 5-second test, per slide: a colleague who has never seen the deck should be able to state the main idea of each slide after 5 seconds of viewing.
• 6×6 audit: open the deck on a single screen and count lines per slide and words per line. Anything over 6/6 needs cutting.
• Back-row test: open the deck at the size shown on the projector and read the bottom row of the smallest table from 6 meters away. If you cannot, the text is too small.
• Greyscale test: convert the deck to greyscale (PowerPoint: View → Greyscale; Keynote: print preview in greyscale) and confirm each slide still tells its story.
• CVD test: pass a few representative slides through a CVD simulator. Distinguishability holds.
• Print test: print the deck 6-up on letter/A4 and read it. If you cannot, the deck is not speaker-ready.
• Cold-read rehearsal: read the speaker notes aloud, slide by slide, at speaking pace, and time it. If the talk runs over by more than 10%, cut.

Open alternatives

References

Internal cross-links to other ors-* skills:

• ors-scientific-visualization-figure-design — DPI, fonts, and multi-panel layout that the slides will use.
• ors-scientific-visualization-schematics-diagrams — workflow and study-design slides.
• ors-scientific-visualization-color-and-accessibility — palette choice and contrast audits.
• ors-scientific-visualization-poster-design — when the same story is told as a poster instead of a talk.
• ors-scientific-writing — caption and headline-as-conclusion phrasing.

External resources (do not fabricate exact paths):

• Beamer user guide (in TeX Live; texdoc beamer or ctan.org/pkg/beamer)
• Marp documentation and themes — marpit.org
• Quarto reveal.js format — quarto.org/docs/presentations/revealjs
• reveal.js documentation — revealjs.com
• WCAG 2.1 contrast guidelines — w3.org/TR/WCAG21/
• WebAIM contrast checker — webaim.org/resources/contrastchecker
• Nature / Science / Cell author guidelines for talks (often in a separate "presentation" section)

Changelog

• 1.0.0 (2026-06-10): Initial adaptation by Pradyumna Jayaram from scientific-slides (K-Dense Inc.). Consolidated talk-demo scripts into a template, added Beamer / Marp / Quarto code patterns, added a results-first talk-time table, dark-mode contrast guidance, and an accessibility checklist.