PLOS Computational Biology Submission Guide: Files, Code, and Fit

PLOS Computational Biology asks for novelty, importance to a field, biological significance, and conclusions justified by the study. That combination distinguishes it from a venue selected only because a manuscript contains computation.

Write two sentences before submission. The first should state what changes computationally: a model, algorithm, inference, representation, resource, or theoretical result. The second should state what changes biologically because of it. If the second sentence merely says the method “may be useful,” the fit case needs more work.

For a new method, compare against credible baselines and explain why the evaluation represents the intended biological job. For a modeling paper, separate fit from prediction and show identifiability or uncertainty where relevant. For software or a resource, document installation, inputs, outputs, versions, tests, and maintenance boundaries.

The journal currently allows initial submission as a single PDF containing text and figures. Accepted manuscript formats include DOC, DOCX, RTF, and PDF; LaTeX manuscripts are submitted as PDFs. The guidelines do not impose a general word, figure, or supporting-information limit, but they explicitly ask for concise presentation.

Format-free does not mean information-free. The PDF still needs the title page, abstract, Author Summary, main sections, methods or models, acknowledgments, references, and supporting-information captions appropriate to the article. Make every figure legible at review size and ensure the PDF contains working cross-references.

At revision, expect to provide separate text, figure, and supporting-information files. Keep editable sources organized from the beginning rather than rebuilding the package after review.

The full title should be specific and concise; the current guidelines cap it at 200 characters and also request a short title. The technical abstract should state the problem, approach, result, and conclusion without overselling the design.

The Author Summary is not a duplicate abstract. It should help a scientist outside the immediate computational specialty understand:

• the biological problem;
• why existing approaches leave an important gap;
• what the study did in plain scientific language;
• what was learned; and
• what the result does and does not enable.

Test the Author Summary with a colleague in a neighboring field. If they can repeat the biological consequence but not the implementation details, it is doing its job.

PLOS requires data underlying the findings to be available without restriction, with rare exceptions, and applies the same principle to author-generated code underlying the findings. The submission includes a Data Availability Statement.

A strong computational package identifies the repository, persistent release or archive, license, software environment, documentation, and exact route from raw or permitted input data to the headline result. Record restrictions honestly where sensitive or controlled data are involved, and follow the journal's exception process rather than promising access that cannot be delivered.

Before upload, ask a colleague to reproduce one central table or figure from the documented materials. Record seeds, model versions, preprocessing, exclusion rules, hyperparameters, and compute assumptions that materially affect the output.

The cover letter is uploaded separately. Current journal guidance asks authors to explain why the manuscript is suitable and why it will inspire the field or drive research forward. Answer those questions with evidence rather than praise.

Editorial Manager then collects article type, files, author information, declarations, and other submission details. PLOS contacts all authors by email at submission. Reconcile names, order, affiliations, and email addresses before the corresponding author begins the form.

The journal's submission guide describes six major form sections and accepts a manuscript plus cover letter, figures, and supporting information, with a combined initial file also available. Preview the compiled PDF before approval.

Use the reporting guideline appropriate to the study design. Human- or animal-research manuscripts need the applicable ethics information in Methods. Check competing interests, funding, author contributions, data availability, software availability, and any preprint information against the portal responses.

For machine-learning work, report the data split, leakage controls, cohort construction, missingness, class balance, external validation, calibration where relevant, and uncertainty. For simulations or mechanistic models, make parameter provenance and sensitivity visible. Avoid a single headline metric when the scientific decision needs several dimensions.

The method wins a benchmark but the biology does not change. A higher predictive score is not automatically biological significance. Identify the biological inference, experiment, or decision that becomes possible. If the contribution remains algorithmic, compare journals that own computational methods without requiring the same biological consequence.

The test set answers an easier question than the manuscript claims. Random splits can leak related samples, subjects, sites, species, time periods, or experimental batches across training and evaluation. Match the split to intended use. If the system will be applied to unseen laboratories, cohorts, organisms, or time periods, evaluate that boundary or limit the claim.

The repository is public but not reproducible. A code link without an archived release, environment, license, instructions, and data route does not let readers reconstruct a result. Create a clean installation, run the central workflow, and record the exact command and expected output. State any controlled-data steps separately.

The Author Summary explains the technology rather than the finding. Replace architecture names and implementation detail with the biological problem, what was learned, and why it matters. Keep one sentence on the approach so readers understand the evidence, but do not turn the summary into marketing copy for the software.

Example: separating prediction from biological explanation

Imagine a model predicts a phenotype from transcriptomic data. A strong predictive result may support classification or prioritization, but it does not by itself identify a causal pathway. The manuscript should distinguish performance, interpretation, and mechanism. Validate important features against perturbation or independent evidence where the biological claim requires it; otherwise describe them as predictive associations.

This distinction should stay consistent in the title, abstract, Author Summary, results headings, and discussion. Reviewers should not encounter a calibrated prediction in Methods and a mechanistic claim in the conclusion.

Treat the manuscript and repository as two views of the same study. Dataset versions, sample counts, exclusions, model settings, and output values should match. Link a persistent release rather than only a moving default branch. Provide a minimal example that lets a reader inspect inputs and outputs without reproducing the full computational cost.

When privacy or licensing prevents raw-data sharing, document the access process and provide code that can operate on permitted or synthetic inputs. A useful restriction statement tells readers who controls access, what criteria apply, and which parts of the result remain independently testable.

• Confirm a computational contribution and a biological consequence.
• Compare the manuscript with recent papers serving the same reader job.
• Prepare a specific full title and short title.
• Separate the technical abstract from the accessible Author Summary.
• Make central code and data available with documentation and licensing.
• Write a precise Data Availability Statement.
• Match reporting and ethics statements to the study design.
• Reconcile author order, affiliations, contributions, funding, and conflicts.
• Answer the cover-letter suitability and field-advance questions directly.
• Inspect the Editorial Manager compiled PDF before approval.

Browse the journal directory, compare PLOS Biology submission requirements, or work through the pre-submission review guide.