eduKateSG · Why Science?
A study becomes easier to inspect when its description, links and supporting files travel together
Use this guide to read a BioStudies package as a map of evidence: identify the study, locate each file or external record, and keep the limits of reuse visible.
BioStudies is useful because a modern biological study rarely fits into one neat file. A paper may point to raw measurements in a specialist archive, processed tables in a supplement, images in another repository and scripts or notes that do not belong in any of those places. The BioStudies database at EMBL-EBI gives those pieces a study-level home: a description, links to external records and files that can be stored with the package. For science learning, this is a practical lesson in provenance. Evidence is easier to understand when you can see how the pieces belong together.
This guide treats BioStudies as a map, not a magic answer. You will practise identifying the study, reading its package structure, following linked records, checking file roles and describing what was actually inspected. The worked examples are fictional and clearly labelled. They are designed for secondary science, science enrichment and early research-methods practice; they are not claims about a real BioStudies accession.
Chapter index
Chapters 1–9: from accession to package structure
- 1. Begin with the study, not the most exciting file
- 2. Treat the accession as the package address
- 3. Read the description as a scope statement
- 4. Sketch the package before downloading it
- 5. Separate hosted files from external links
- 6. Use sections as organisation, not biological proof
- 7. Let file lists reveal the practical dataset
- 8. Follow specialist archive links for structured detail
- 9. Distinguish public, private and incomplete access
Chapters 10–17: from provenance to responsible reuse
- 10. Record provenance at the level you used
- 11. Inspect formats before interpreting values
- 12. Read citation and reuse instructions separately
- 13. Check integrity without pretending it proves validity
- 14. Work a fictional package from question to file
- 15. Compare packages with a declared question
- 16. Turn the package into a reproducible school report
- 17. Connect study packages to science learning and careers
CHAPTER 1 OF 17 · Find the study package
1. Begin with the study, not the most exciting file
A BioStudies package has a study-level identity. That matters because a spreadsheet or image detached from its study can look convincing while losing the question, organism, treatment, time point and processing history that make it interpretable. Start by reading the title, accession, description and any publication link. Ask what the study says it contains. Then compare that promise with the visible sections, files and external links. A package is a route through evidence, not a guarantee that every item answers your question.
Try this
Imagine a fictional study called “Seedling Light Shift”. Its package contains a short description, a table of sample labels, photographs and a link to raw sequencing data in a specialist archive. Before opening `final_results.xlsx`, write a one-sentence package question: “Which files show the treatment design, and which files contain observations?” This keeps the attractive result table from becoming the whole story.
Record four items in your notebook: study accession, biological question, main organism or material, and the kinds of supporting objects listed. If one is missing, write “not established from the package summary” rather than filling the gap from memory. That small habit turns browsing into accountable science learning.
Evidence checkpoint
State the study question and list the package objects you actually inspected. A result file without its design context cannot carry the whole conclusion.
CHAPTER 2 OF 17 · Find the study package
2. Treat the accession as the package address
An accession is a stable route to a record, but it is not a scientific conclusion. Copy it exactly, including prefixes and punctuation, and keep it beside the title. When two browser tabs look similar, the accession helps you notice whether you are reading the same study, a collection, a related archive record or a later package. Citations and laboratory notes become much easier to audit when the identifier travels with every claim derived from the package.
Try this
For the fictional Seedling Light Shift example, suppose your teacher gives you `S-STUDY-204`. Write the accession above your notes and add the access date. Now imagine a classmate sends a screenshot labelled only “Figure 2”. Ask for the accession and file path before comparing it with your copy. The goal is not bureaucracy; it is to ensure both of you are looking at the same evidence object.
Practise an identifier check with three invented strings: `S-STUDY-204`, `E-PLANT-204` and `Figure2.png`. Classify the first as the study package address, the second as a possible external archive address and the third as a file name. The classifications are fictional, but the reasoning is reusable: package, linked record and file occupy different levels.
Evidence checkpoint
Copy the package accession, any linked accession and the file name into separate fields. This prevents a study, external record and file from being cited as though they were the same object.
CHAPTER 3 OF 17 · Find the study package
3. Read the description as a scope statement
The study description should tell you enough to decide whether the package deserves closer inspection. Look for the research question, material, design and broad data types. Do not assume that a short description contains every method or limitation. Use it to orient yourself, then look for the files, publication or archive record that carries the detail. A description that says “response to light” does not establish which wavelengths, exposure durations or outcome measures were used.
Try this
In our fictional package, the description says seedlings were moved from low light to high light and sampled over time. That supports a broad time-course interpretation. It does not yet tell you whether the same individuals were followed, whether different seedlings were harvested at each time point, or how high light was defined. Write those as open questions. Open questions are part of evidence reading, not signs that you have failed.
Turn a vague description into a scope grid with four headings: stated, linked, missing and inferred. Put only exact package statements under “stated”. Put destinations such as a sequencing accession under “linked”. Put unresolved design details under “missing”. Reserve “inferred” for your cautious hypotheses and label them clearly. This grid prevents an inference from quietly becoming a database fact.
Evidence checkpoint
Underline every word supplied by the study description. Circle every word you inferred. If the circles dominate, return to the files or publication before writing a conclusion.
A package can contain sections, files and links. Sketching that structure first reduces confusion when file names repeat across folders or when one archive record contains many samples. Draw the study as a central box. Add one branch for each section. Under each branch, list hosted files and external links separately. This simple diagram becomes a reading plan and exposes empty-looking branches that may be explained elsewhere.
Try this
Suppose Seedling Light Shift has sections named “Design”, “Images” and “Molecular data”. Design contains `samples.tsv`; Images contains twelve JPEG files; Molecular data links to a specialist archive. Your sketch immediately shows that the raw molecular measurements are not necessarily inside the BioStudies package. It also stops you from describing the JPEG folder as the entire dataset.
After sketching, mark each object with a purpose word: identity, design, raw, processed, figure, code, protocol or correspondence. If the purpose is unclear, use a question mark. Do not decide from the extension alone. A CSV can be a sample sheet, a processed result or a software export; its role comes from context and documentation.
Evidence checkpoint
Your sketch should show at least one branch for hosted material and one for external destinations when both are present. Label unknown roles with a question mark.
BioStudies can hold files that do not fit a structured archive and can link to data held elsewhere. These are complementary roles. A hosted supplementary table is retrieved from the package, while an external accession asks you to inspect another database with its own metadata and rules. When reporting what you checked, say whether you opened the hosted object or followed the external record. “Available through BioStudies” is too vague if the evidence actually lives in another archive.
Try this
In the fictional package, `samples.tsv` is hosted, while raw sequence reads are represented by an external link. If you inspect only the sample table, you can describe labels and planned comparisons. You cannot claim to have checked read quality. If the external link fails, record that access problem rather than treating the sample table as a substitute for the raw data.
Make a two-column list called “inside package” and “linked elsewhere”. For each item, add the action needed to verify it: open file, inspect archive record, read method, or request access. This makes the next step visible and is especially useful in group work, where different students can inspect different layers without duplicating effort.
Evidence checkpoint
Report whether you opened a hosted file or followed an external link. The phrase ‘available through BioStudies’ is too broad when the evidence lives elsewhere.
CHAPTER 6 OF 17 · Trace files and links
6. Use sections as organisation, not biological proof
Sections make a complex package readable. They may group figures, assays, supplementary material or administrative files. The grouping helps navigation, but the section label does not itself validate the content. A folder called “controls” does not prove that the controls were appropriate. A section called “raw data” still needs file-level inspection. Keep organisational meaning separate from evidential meaning.
Try this
For Seedling Light Shift, imagine a section labelled “Validation”. It contains a PDF describing a second measurement and a small results table. The label suggests why the authors grouped the material. To evaluate validation, you still need to ask what was measured, whether the samples were independent and whether the result addresses the original claim.
Choose one section and write two sentences: “The section groups…” and “To judge the evidence, I still need…”. This pair is a useful antidote to confident folder names. It also helps a tutor or teacher see whether you understand the difference between a package’s structure and the scientific reasoning supported by its contents.
Evidence checkpoint
A section name explains organisation. To judge validity, identify the experiment, sample, method or document inside that section that supports the claim.
CHAPTER 7 OF 17 · Trace files and links
7. Let file lists reveal the practical dataset
A file list tells you what can actually be retrieved. Read names, extensions, paths and sizes as clues, then confirm their roles from documentation. Look for a sample sheet, a readme, checksums, code, processed results and figure source data. Repeated prefixes may show batches or conditions. Missing documentation may be a limitation even when the files are present.
Try this
In our fictional example, twelve images use names such as `L2_T03_R1.jpg`. You hypothesise that the code means light condition 2, time 3 and replicate 1. Do not publish that decoding until a key confirms it. Write the pattern you observe and the documentation you used. If the package includes `README.txt`, read it before sorting the images.
Create a five-row file inventory. For every row, record path, format, likely role, confirmation source and whether you opened it. This is more informative than a download folder full of unexplained files. It also supports careful science tuition: the student practises classification, evidence and uncertainty with real research infrastructure rather than memorising database names.
Evidence checkpoint
Confirm a filename pattern with a readme or sample sheet before decoding it. Repeated characters are clues, not automatically documented metadata.
CHAPTER 8 OF 17 · Trace files and links
8. Follow specialist archive links for structured detail
Specialist archives often describe a data type more precisely than a general study package can. A sequencing archive may expose samples and runs; an expression resource may show experimental factors; an image archive may describe imaging datasets. BioStudies can connect these records at the study level. Follow the link and keep both identifiers so that the relationship remains visible.
Try this
Suppose the fictional package links to an expression accession. On the destination record, inspect sample count, organism, assay and factor labels. Then return to the study package and ask whether its description and sample sheet agree. A link is useful because it connects records; it does not remove the need to compare their metadata.
Write a bridge sentence with both identifiers: “Study package X links to archive record Y for the raw molecular measurements.” Then add what you verified at Y. Avoid the shortcut “BioStudies contains the sequencing data” unless you actually confirmed where the files are stored. Accurate location language helps another reader reproduce your route.
Evidence checkpoint
Keep the study accession and specialist-archive accession together in your notes. Record which metadata agreed and which fields still differed or remained unclear.
CHAPTER 9 OF 17 · Check reuse and provenance
9. Distinguish public, private and incomplete access
Submission systems may allow authors to inspect material before public release, and some packages can contain links or files with access conditions. A secret submission link, an embargoed object and a public accession are different access states. Never share a confidential link, and never describe material as public because you personally can open it. Record the access state you observed and the date.
Try this
For a classroom exercise, imagine that `methods-v3.pdf` appears in a private preview but not on the public package. Your notes should say that the public record did not expose the file at the time checked. Do not use the preview URL in a report or send it to classmates. The science lesson is simple: access is part of provenance.
On any package, look for release status, access instructions and visible warnings. If access fails, capture the accession and the ordinary public URL, not authentication tokens. A broken or restricted link is a finding about accessibility, not evidence that the underlying experiment did or did not happen.
Evidence checkpoint
Describe access as public, restricted, preview-only or unavailable at the time checked. Never copy an authentication or secret-preview token into shared notes.
CHAPTER 10 OF 17 · Check reuse and provenance
10. Record provenance at the level you used
Provenance answers where an object came from and how it reached your analysis. At minimum, record the study accession, file path or linked accession, access date and any visible version or release information. If you transform a file, keep the command or steps and write a new output name. This lets a future reader separate archive content from your work.
Try this
In Seedling Light Shift, imagine you filter `samples.tsv` to six high-light rows and save `samples_high_light_student.csv`. Your file is not an archive original. Put a header in your notes: “Derived from S-STUDY-204/samples.tsv; filtered by condition; 24 October 2026.” The fictional accession is only an exercise, but the provenance pattern is ready for real use.
Practise a three-generation chain: source file, cleaned file, summary figure. Give each a distinct name and one-line transformation note. If you cannot reconstruct the chain, pause before interpreting the figure. Good data handling is part of scientific reasoning and increasingly relevant to career pathways in laboratory science, bioinformatics and research data management.
Evidence checkpoint
Name every transformation you performed after download. A filtered table and an archive original must never share the same filename or citation description.
CHAPTER 11 OF 17 · Check reuse and provenance
11. Inspect formats before interpreting values
File format shapes what you can safely read. Plain text tables need a delimiter and header; spreadsheets may contain hidden sheets or formulas; images need scale and channel information; compressed archives need an inventory; PDFs can mix selectable text with embedded figures. Identify the format and choose an appropriate viewer before judging content.
Try this
Our fictional sample table has tab-separated columns. If a spreadsheet program silently converts an identifier into a date, the visual table may look tidy while the identity changes. Open a copy as plain text, check the header and preserve the original. For an image, seek the scale, acquisition context and any associated legend instead of measuring pixels as if they were automatically biological units.
Write a format check for one object: “I opened X with Y; I confirmed Z; this view cannot establish Q.” That final clause is powerful. It prevents the convenience of a viewer from being mistaken for completeness of evidence.
Evidence checkpoint
Check delimiter, headers, formulas, units and encoding before interpreting numbers. A viewer can change appearance without changing the underlying file—and sometimes can silently change values.
CHAPTER 12 OF 17 · Check reuse and provenance
12. Read citation and reuse instructions separately
A public page is not the same as unrestricted reuse. Look for citation guidance, licences and third-party notices. Cite the study package when you use its organisation or hosted material, and also cite the linked specialist record or publication when your claim depends on it. If a file has its own licence, follow that licence rather than assuming the whole package has one rule.
Try this
In the fictional example, suppose the photographs have a creator notice while the sample table has a permissive data licence. You may be able to recompute a summary from the table without being free to republish the photographs. Record the distinction. Attribution should point to the object that supports the statement, not merely the nearest homepage.
Make a reuse checklist with creator, citation, licence, third-party content and modifications. If one field is unclear, link to the source and ask before redistribution. This habit keeps school projects, presentations and later research work both useful and respectful.
Evidence checkpoint
Separate permission to read from permission to redistribute. Record the licence and any third-party notice for the specific object you plan to reuse.
CHAPTER 13 OF 17 · Explain the evidence clearly
13. Check integrity without pretending it proves validity
Checksums or file sizes can help confirm that a downloaded file matches the distributed object. They do not prove that the experiment was well designed or that the analysis was correct. Integrity, documentation and scientific validity are related but separate checks. Name the check you actually performed.
Try this
Suppose a package publishes a checksum for `samples.tsv`. Matching it after download supports file identity. It does not tell you whether sample labels were entered correctly. A mismatch may indicate an incomplete transfer or a different version, so retrieve the file again and compare the package information before proceeding.
Write three headings in your notebook: “same file?”, “understandable file?” and “convincing evidence?”. Place checksum under the first, readme and headers under the second, and design, controls and analysis under the third. This prevents a technical pass from turning into an oversized scientific claim.
Evidence checkpoint
A checksum can support file identity. It cannot establish correct sample labels, suitable controls or a valid biological interpretation.
CHAPTER 14 OF 17 · Explain the evidence clearly
14. Work a fictional package from question to file
Now combine the method. The fictional question is: “Did the light shift change seedling colour by the final time point?” The package sketch shows a design table, photographs and a linked molecular dataset. The colour question points first to images and their acquisition notes, not automatically to the molecular archive. You locate the image section, confirm time labels from the readme and identify the final-time files.
Try this
You compare only equally lit photographs with the same scale and note that image appearance can be affected by capture settings. You record a visual pattern as an observation, not a quantified pigment result. The linked expression record might help explore mechanism later, but it does not replace a direct colour measurement.
Write the conclusion in three layers: “The package contains…”, “Within the inspected images, I observed…”, and “This does not establish…”. This structure keeps package fact, personal observation and limit distinct. It is a cheerful way to make a modest result genuinely trustworthy.
Evidence checkpoint
Write one sentence for package fact, one for your observation and one for the limit. Do not compress all three into a confident causal claim.
CHAPTER 15 OF 17 · Explain the evidence clearly
15. Compare packages with a declared question
Two BioStudies packages may organise similar material differently. Compare them only after choosing a question. If the question is discoverability, compare descriptions and file inventories. If it is reuse readiness, compare documentation, formats, access and provenance. If it is biological evidence, inspect the underlying study designs rather than ranking packages by file count.
Try this
Imagine Package A has fifty files and a clear readme; Package B has five files plus a well-described external archive link. “More files” does not mean “better evidence”. Package B may simply delegate raw data to the appropriate specialist archive. The comparison should reward clarity and fitness for purpose, not bulk.
Build a rubric with four criteria and a short justification for each score. Keep “not checked” available. A transparent incomplete comparison is stronger than a complete-looking table filled with guesses.
Evidence checkpoint
Choose the comparison question before assigning scores. File count, package size and visual polish are not universal measures of evidential quality.
CHAPTER 16 OF 17 · Explain the evidence clearly
16. Turn the package into a reproducible school report
A strong report gives another student a route. Name the accession, state the question, list the files or linked records inspected and describe any transformations. Put screenshots in a supporting role; use text identifiers and URLs so the route survives a changed layout. Separate database descriptions from your interpretation.
Try this
For a science enrichment assignment, write a compact methods box: “Accessed BioStudies on 24 October 2026; opened the public study page; inspected the description, file list and linked archive record; downloaded no restricted material; calculated no new biological statistic.” Adjust it to match what you actually did.
Ask a partner to reproduce one step using only your report. If they cannot find the same object, improve the identifier or path. Reproducibility can begin with a five-minute navigation check long before advanced laboratory work.
Evidence checkpoint
Give a classmate enough information to reproduce one navigation step. If the route depends on your screenshot alone, add the accession, path and ordinary public URL.
CHAPTER 17 OF 17 · Explain the evidence clearly
17. Connect study packages to science learning and careers
BioStudies makes an important idea visible: scientific evidence is also an information system. Researchers need people who can curate metadata, maintain archives, design submission workflows, check formats and help others find the right object. These tasks connect classroom skills—careful reading, classification, tables and honest uncertainty—to careers in research software, data stewardship, librarianship, bioinformatics and laboratory management.
Try this
The next step is small. Choose one public study package related to a topic you already understand. Draw its map, inspect one hosted file and one external link, and write a three-sentence scope statement. Do not try to analyse everything. A well-traced small route teaches more than a fast download of hundreds of unexplained files.
Continue with the Science Learning Hub for subject routes, the Education Hub for wider learning choices, and the careers-by-subject guide when you want to see how evidence-handling skills travel into work.
Evidence checkpoint
Name the capability you practised—curation, provenance, format checking or evidence writing—and connect it to one next learning step rather than a guaranteed career outcome.
Contents · Previous chapter · Continue at the Science Learning Hub
A calm five-minute package check
Before leaving a BioStudies record, confirm five things: the study accession, the package description, one hosted object, one linked destination when present, and the reuse or citation guidance for the material you plan to use. Then write a limit beside the result. This small routine makes a large package easier to revisit and gives another reader a route they can actually follow.
| Reading object | What it can support | Question to keep visible |
|---|---|---|
| Study description | Identity, scope and high-level context | Does the description match the question I am asking? |
| Hosted file | Material stored with the study | What format, version and role does this file have? |
| External link | A route to data held in another archive | Does the destination resolve to the intended record? |
| Section or file list | Organisation of related material | Is the grouping author-supplied, archive-generated or my own? |
Sources and next reading
Substantive resource claims in this article were checked against the primary pages below. A database record is a structured route into evidence; it does not by itself prove a new biological conclusion. Recheck the live record, evidence labels and release context before using it in assessed work or research.
Primary sources
- EMBL-EBI BioStudies home and scope — accessed 24 October 2026.
- BioStudies submission help — accessed 24 October 2026.
- BioStudies ArrayExpress collection help — accessed 24 October 2026.
- EMBL-EBI announcement: BioStudies as a study data package — accessed 24 October 2026.