Help & Documentation

CircadiOmics is a web portal that aggregates circadian multi-omics datasets (transcriptome, metabolome, proteome, acetylome) in both bulk and single-cell modalities and provides interactive visualization together with statistical tests for rhythmicity using BIO_CYCLE.

Topics

Selecting Datasets

Datasets in CircadiOmics are grouped by omic type. The main page provides four dropdown menus — Transcriptome, Metabolome, Proteome, and Acetylome — each listing the datasets currently available for that omic.

Click an entry in any dropdown to add it to the "Selected Datasets" panel. You can add as many datasets as you like from the same or different omic types; every subsequent query will be run against every dataset currently in this panel. This makes it easy to compare the same gene or metabolite across tissues, conditions, or species in a single display.

Interacting with entries in the "Selected Datasets" panel:

  • Click the yellow dataset name to open a pop-up with the full citation, species, omic type, experimental condition, timepoints, and any sibling datasets from the same study.
  • Click the red X button to remove a dataset from the selection. Queries will no longer be run against it.

If you only want to browse metadata without running any queries, use the Datasets page instead — see Dataset Details below.

Dataset Details

The Datasets page provides the main metadata reference for datasets available in the portal. It is reachable from any page via the "Datasets" button in the top navigation bar.

The page is organized into four tabs, one per omic type (Transcriptome, Metabolome, Proteome, Acetylome). Depending on the dataset, the table may include:

  • A short human-readable description of the study.
  • Species (e.g., Mus musculus, Homo sapiens, Drosophila melanogaster).
  • Tissue or organ (e.g., Liver, Brain, Heart, Suprachiasmatic Nucleus).
  • Experimental condition (e.g., Control, Knockout, Knockdown, Drug Treatment, Ad Libitum Feeding, or Restricted Feeding).
  • A direct link to the original publication (DOI or PubMed).
  • The timepoints sampled in the experiment.

A keyword search box above the tables lets you filter rows by any of the columns simultaneously — for example, typing liver surfaces every liver dataset, regardless of omic type or species.

Searching Molecular Species

After selecting one or more datasets, type a molecular identifier such as a gene symbol, metabolite name, protein ID, or other measured feature in the main search bar. The search bar provides autocomplete suggestions when matching identifiers are available; use the arrow keys to navigate suggestions and Enter to accept one.

Searches are case-insensitive, so Per1, PER1, and per1 are treated as the same query. If the queried identifier is not present in a selected dataset, that dataset is skipped in the display.

Running and Clearing Queries

  • Click Display to add the current query to the plot and statistics table. The display supports multiple molecular identifier × dataset traces at the same time.
  • Click Clear to remove every trace and reset both the plot and the statistics table. "Clear" does not remove selected datasets; it only clears query results.

Performing Batch Queries

Batch queries let you explore how many molecular identifiers behave across many datasets simultaneously. There are two independent axes you can batch along.

Dataset Batch

To compare a single symbol across many datasets, add each dataset of interest to the "Selected Datasets" panel before running the query. Any combination of datasets — regardless of species, tissue, or omic type — can be selected at the same time. See Selecting Datasets for details.

Example use case: to compare Bmal1 oscillations across mouse liver, heart, and SCN, add the three datasets to "Selected Datasets" and query Bmal1 once.

Symbol Batch

To query multiple symbols against the currently selected datasets at once, separate each symbol with a semicolon (;) in the search bar. For example: 

Per1; Per2; Cry1; Bmal1

Each symbol is run against every selected dataset, so a symbol batch combined with a dataset batch yields a full matrix of traces. Symbol lookup is case-insensitive, and whitespace around the semicolons is ignored.

Filtering Datasets

The dataset dropdowns include a filter menu so you can narrow down the list before picking. Filters are grouped into three categories:

  • Species — e.g., Mus musculus, Homo sapiens, Drosophila melanogaster, Rattus norvegicus.
  • Tissue — broad tissue categories such as Liver, Muscle, Brain, Heart, Adipose, Skin, and others.
  • Experimental Type — classification of the perturbation, e.g., Control, Knock-Out, Knock-Down, Overexpression, Drug Treatment, Dietary intervention.

Multiple filters can be active at the same time, and filters within a category are combined with a logical OR (show datasets matching any of the selected values), while filters across categories are combined with AND (show datasets matching at least one value in every chosen category).

The "Clear Filters" option at the very top of the filter menu resets every filter and returns the dataset dropdown to its full contents.

Statistics

Each row of the statistics table corresponds to one symbol × dataset combination. Values are produced by whichever software is currently selected in the "Software" toggle (see Displaying Results).

P-Value

The p-value reflects the statistical evidence against a non-rhythmic signal, as reported by the selected software. Lower values indicate stronger evidence for oscillation. By convention, a p-value < 0.05 is often treated as statistically significant, but this threshold should be adjusted for the number of identifiers being tested — consider the Q-value below for multiple-testing correction.

Q-Value

The Benjamini–Hochberg False Discovery Rate (FDR) corresponding to the P-value, as reported by BIO_CYCLE or JTK_CYCLE. A Q-value threshold of 0.05 controls the expected false discovery rate at approximately 5% among features called significant, under the assumptions of the correction procedure. Q-values are the preferred criterion when screening large numbers of identifiers for rhythmicity.

Amplitude

The estimated strength or size of the rhythmic oscillation, reported in the same units as the input measurements (e.g., log-expression or normalized intensity). For sinusoidal fits, amplitude is commonly interpreted as half the peak-to-trough distance. Amplitude can be compared between identifiers measured on the same scale, but should not be compared directly between datasets with different normalization.

Lag (Phase)

The phase shift of the fitted periodic signal, typically expressed in the same units as the experimental timepoints (hours). Lag describes when during the cycle an identifier peaks, which is often more biologically informative than amplitude alone.

Period

The period length of the fitted signal. BIO_CYCLE estimates the best-fitting period from the candidate period range supported by the model and portal configuration; JTK_CYCLE evaluates a user-supplied discrete set of candidate periods (commonly 8, 12, and 24 h in the portal-wide scan).

Displaying Results

Removing Individual Traces

Every row of the statistics table has a "Remove" button; clicking it removes that specific symbol × dataset pair from both the plot and the statistics table, while leaving other traces untouched. This is useful when one trace has a very different scale and is compressing the rest of the display.

Normalization

The "Normalization" toggle switches between:

  • Off — raw expression/concentration values as reported by the original publication. Use this to inspect absolute magnitudes.
  • On — each trace is rescaled to the [0, 1] range. Use this when comparing shape and phase across symbols or datasets whose absolute scales differ substantially.

The statistics table is unaffected by the Normalization toggle — p-values, amplitudes, and phases are always computed from the raw input data.

Software

The "Software" toggle switches the statistics table between BIO_CYCLE and JTK_CYCLE outputs. The plotted time-courses do not change — only the statistics columns are refreshed to reflect the selected software. See Statistics for field definitions, and BIO_CYCLE Results for method notes.

Downloadable Content

CircadiOmics lets you download every artifact displayed in the portal so you can reproduce figures and continue analysis offline.

Plot Images

The current display can be exported as PNG (raster) or SVG (vector) using the buttons inside the query search bar. SVG is recommended for figures intended for publication, as it scales without loss of quality and can be edited in vector-graphics software.

Statistics Table

The statistics table can be exported as an Excel-compatible file with the "Excel" button on the upper right of the table. The export reflects the currently selected Software toggle, so switch to the software of interest before clicking Excel if you want BIO_CYCLE vs. JTK_CYCLE values.

Processing Data with BIO_CYCLE

The BIO_CYCLE portal lets you upload your own time-course measurements and obtain the same rhythmicity statistics described in Statistics. The same tool that powers CircadiOmics' built-in statistics is applied to your data, so results are directly comparable.

Step-by-Step Instructions

  1. Click "BROWSE FILES" and select your .tsv dataset. A green "File selected — ready to run BIO_CYCLE" confirmation will appear below the buttons. See the Input File Format section below or download the example file for the correct layout.
  2. Once you see the green confirmation, click "RUN BIO_CYCLE" (next to the Browse button). A progress bar and status message will appear. When the file has finished uploading you will see an "Upload complete!" message, at which point BIO_CYCLE analysis starts automatically across all three candidate periods (8, 12, and 24 h). When analysis finishes, a results file containing p-value, q-value, period, lag, and amplitude for each row downloads automatically.
  3. Two view buttons appear after a successful run — choose Histograms or Visualization.
  4. Histograms — examine the distribution of statistics across all molecular species:
    1. Select a statistical metric: P-Value or Q-Value.
    2. Select a significance threshold: 0.05, 0.01, or 0.001.
    3. Select a period: 24, 12, or 8 hours.
    4. Click Lag, Amplitude, or Offset to display the histogram for that statistic, filtered to species that pass the chosen threshold.
    5. Click IDs to list the specific species that meet the threshold, or Download to export them as a zip archive.
  5. Visualization — plot the time-course trajectory of a single molecular species:
    1. Type the species ID in the search box. Suggestions from your uploaded dataset appear as you type.
    2. Click Display to render the expression time series.

Output Tabs

After a successful upload, the BIO_CYCLE portal exposes:

  • Visualization — interactive per-symbol time-course traces for any symbol in your uploaded file.
  • Histogram — distributions of the main statistics (P-value, Q-value, Amplitude, Period, Lag) across all symbols, useful for an at-a-glance quality check.

Statistics are computed for every symbol in the uploaded file across three candidate periods (8, 12, and 24 h) by default.

Input File Format

Uploaded data must be a tab-separated file (*.tsv) with:

  • A single header row.
  • A first column named ID containing unique symbol IDs.
  • One column per measurement, named ZT_<time>_REP_<n> (for Zeitgeber time) or CT_<time>_REP_<n> (for circadian time). The time portion is in hours. Replicates are 1-indexed.

BIO_CYCLE tolerates:

  • Variable numbers of replicates per timepoint.
  • Uneven timepoint spacing.
  • Missing replicates (write na for any missing value). Every row must have at least one numeric value per timepoint.

If you need a starting template, download the example below and edit it in your preferred spreadsheet tool. Make sure to save as tab-separated (.tsv), not comma-separated (.csv).

Download example BIO_CYCLE input file

ID ZT_0_REP_1 ZT_0_REP_2 ZT_4_REP_1 ZT_4_REP_2
Per1 4.25 3.08 9.17 8.90
Per2 3.45 na 9.50 9.93
Per3 1.40 1.53 10.70 10.55
Cry1 4.35 5.78 14.87 12.21

BIO_CYCLE Results

BIO_CYCLE is a deep-learning-based method for detecting rhythmicity in biological time-course data. It is trained on synthetic periodic and non-periodic signals designed to mimic the noise and sampling characteristics of real circadian experiments, and reports p-values, q-values, period, amplitude, and phase estimates per identifier.

Because BIO_CYCLE can evaluate candidate periods beyond a fixed 24-h rhythm, it can identify oscillatory patterns that do not fall exactly on the classical 24-h frequency — a common situation in circadian experiments under free-running conditions, in non-mammalian model organisms, or in aging/stress contexts.

BIO_CYCLE is the default method used throughout CircadiOmics. JTK_CYCLE statistics are also provided where available and can be accessed via the Software toggle; see Displaying Results. For the meaning of each statistic column, see Statistics. To run BIO_CYCLE on your own data, see Processing Data with BIO_CYCLE.

Single-Cell Data and Annotations

The Single-Cell section brings circadian analysis to cell-type resolution. For every single-cell dataset, cells are either annotated using a reference atlas (when the original study did not provide annotations) or displayed with the authors' own annotations, and BIO_CYCLE statistics are computed per cell type.

The Dataset Table

The "Available Datasets" table at the top of the page lists every single-cell study currently integrated in the portal, with the following columns:

  • Dataset Name — a human-readable label (e.g., "Mouse Aorta Light-Dark") plus a short description of the study.
  • Species — source organism.
  • Tissue — tissue or organ of origin.
  • Timepoints — sampling timepoints used in the experiment.
  • Paper — link to the associated publication.

Interactive Exploration

Below the table, the "Interactive Gene Expression Analysis" panel lets you explore any single gene in any cell type of any dataset:

  1. Pick a Dataset in the first dropdown.
  2. Pick a Cell Type in the second dropdown.
  3. Type a Gene name; an autocomplete dropdown shows matching genes. Use ↑/↓ and Enter to pick a suggestion, or Esc to dismiss.
  4. Click Display. The plot below shows mean expression per timepoint with standard-error bars, and the Results Table reports BIO_CYCLE statistics for every gene currently selected.

You can add more genes one at a time; each additional gene is overlaid on the same plot and appears as a new row in the Results Table. Individual genes can be removed from the display using the Remove button in their table row. Use Clear to reset the entire panel.

What is in the Downloads?

Each dataset exposes up to two downloads:

  • BIO_CYCLE results archive (<Dataset_Title>_biocycle_results.tar.gz) — a per-study archive containing one folder per cell type. Each folder holds a .tsv file with BIO_CYCLE statistics (ID, P-value, Q-value, Period, Amplitude, Lag) and one column per timepoint replicate (TP_<time>_<rep>) used to compute those statistics.
  • Annotated single-cell data (.h5ad, or annotated.tar.gz for larger bundles) — an AnnData object containing the expression matrix, cell-level annotations (including predicted cell type and confidence), and dimensionality-reduction embeddings. The .tar.gz bundle additionally includes .png images of the reference UMAPs shown in the interactive panel.

Annotation Policy

For studies that released their own annotations, CircadiOmics preserves and redistributes those labels directly, without re-annotating. For unannotated studies, cells are labeled against the most appropriate publicly available reference atlas for the species and tissue (e.g., Tabula Muris Senis for mouse, Tabula Sapiens for human, the rosbashlab clock-neuron atlas for Drosophila). The atlas used for each dataset is indicated in the Atlas column of the dataset table.

Metabolic Atlas

The Metabolic Atlas is an in-development companion to the Metabolome section of CircadiOmics. Its goal is to identify correlated temporal patterns across the metabolite measurements in the portal and expose them as navigable networks, so that users can identify co-oscillating modules of metabolites sharing a tissue, a species, or a pathway context.

Features and layouts on the Metabolic Atlas page are subject to change as the underlying network reconstruction matures, and some features may be unavailable while this section is under active development. Please cite the corresponding publication (see Citing CircadiOmics) for the specific version that was used in your analysis.

Citing CircadiOmics

If CircadiOmics or its statistics contributed to a publication, please cite both the portal and the BIO_CYCLE software.

  • CircadiOmics: CircadiOmics: Circadian Omic Data Web Portal. M. Samad, F. Agostinelli, T. Sato, K. Shimaji, P. Baldi. Nucleic Acids Research, vol. 50, no. W1, 2022, https://doi.org/10.1093/nar/gkac419.
  • BIO_CYCLE: What time is it? Deep learning approaches for circadian rhythms. F. Agostinelli, N. Ceglia, B. Shahbaba, P. Sassone-Corsi, and P. Baldi. Bioinformatics, vol. 32, no. 19, 2016, pp. 3051-3051., https://doi.org/10.1093/bioinformatics/btw504.
  • For every dataset used, please also cite the original publication linked from the Datasets or Single-Cell tables.

Frequently Asked Questions

Why does a symbol I expect to find not appear in a given dataset?

CircadiOmics only indexes the symbols measured in the original study. If a gene was below the detection threshold, was not profiled on the array, or was filtered out by the authors' pipeline, it will not be queryable for that dataset. Try a different isoform name or an orthologous symbol if you are cross-species comparing.

My symbol shows up in one dataset but not another. Is that a bug?

No — different studies measure different molecules and use different annotation conventions. Batch queries handle this gracefully: datasets that lack the symbol are simply skipped in the plot and the statistics table.

Amplitudes look inconsistent across datasets — why?

Published datasets differ in their normalization pipelines (raw counts, TPM, RMA, log-transformed intensities, etc.), so amplitudes are in different units across studies. Use the Normalization toggle to compare shapes and phases; do not directly compare raw amplitudes across datasets.

How do I export only the rows I care about from the statistics table?

Use the Remove button on rows you do not need, then click Excel. The exported file contains exactly the rows currently visible in the table.

Troubleshooting

  • Nothing happens when I click Display. Make sure you have at least one dataset in the "Selected Datasets" panel and at least one symbol typed in the search bar. Browser extensions that block third-party scripts such as Google Charts or jQuery can also prevent rendering; try again in a private/incognito window.
  • BIO_CYCLE upload rejects my file. Verify that the file is tab-separated (.tsv), the first column is named ID, the header uses ZT_<t>_REP_<n> or CT_<t>_REP_<n> exactly, and that every row has at least one numeric value per timepoint. See the example file.
  • The downloaded archive name collides when I download many datasets. The BIO_CYCLE archive is named after the dataset title, so each dataset saves under a unique filename. If your browser is saving them all under the same name, check that your browser is not overriding the server-supplied filename.

Further Help

If this page does not answer your question, please reach out:

  • Email the CircadiOmics maintainers at kchangiz@uci.edu.
  • For bug reports, it helps to include the URL of the page you were on, the molecular identifier(s) and dataset(s) you had selected, the browser and operating system you were using, and a screenshot of the issue.
  • For data-access requests or questions about specific publications beyond what the portal exposes, please contact the corresponding authors of the original study (linked from the Datasets page).