Covary protocol portfolio and 3.0.1 beta are now available.
Legal

Disclaimer

Please read this disclaimer carefully before using Covary, its associated tools, or any content published on this website. By accessing or using Covary, you acknowledge and agree to the terms set out below.

Research use only Not for clinical or diagnostic use Provided as-is

1. General disclaimer

Covary is a research-grade computational framework developed and maintained by Marvin De los Santos as part of the ChordexBio project. It is made available to the public for research, academic, and educational purposes only.

The information, tools, outputs, protocols, and materials provided through Covary and this website are offered as-is, without warranty of any kind — express or implied — including but not limited to warranties of accuracy, completeness, fitness for a particular purpose, or non-infringement.

Use of Covary is entirely at your own risk. The developer and associated parties shall not be held responsible for any outcomes, decisions, actions, or consequences arising directly or indirectly from the use of this tool or its outputs.


2. Non-clinical and non-diagnostic use

Critical limitation

Covary is not a clinical tool, not a diagnostic instrument, and not approved, validated, or intended for use in medical decision-making, patient care, clinical trials, or regulatory submissions of any kind.

Covary outputs — including phylogenetic trees, embeddings, distance matrices, cluster assignments, and any downstream predictions — are computational approximations for research exploration only. They do not constitute medical advice, diagnosis, prognosis, or treatment recommendations.

Results produced by Covary must not be used to:

  • Diagnose, treat, cure, or prevent any disease or medical condition in humans or animals
  • Guide clinical decisions or patient management
  • Support regulatory filings, IVD submissions, or clinical laboratory reporting
  • Serve as the sole basis for public health interventions or epidemiological policy
  • Replace laboratory-validated methods or certified diagnostic platforms

Any application of Covary outputs in a clinical or regulatory context is the sole responsibility of the user and must comply with applicable laws, institutional review requirements, and professional standards.


3. Non-standardized and experimental nature

Covary is an experimental, research-grade framework. It has not undergone standardized laboratory validation, proficiency testing, or accreditation by any national or international body (e.g., ISO, CAP, CLIA, CE-IVD, FDA).

Results may vary depending on:

  • Input sequence quality, length, and completeness
  • Organism type, genomic region targeted, and level of taxonomic divergence
  • Parameter settings chosen by the user (k-mer size, perplexity, linkage method, etc.)
  • Computational environment (RAM, runtime, Google Colab tier)
  • Covary version used — results may differ between releases

Users are responsible for independently evaluating and validating any outputs before applying them in research publications, grant applications, or downstream analyses.


4. Scope and specificity limitations

Covary was designed and validated primarily for alignment-free, translation-aware phylogenetic and phylogenomic analysis of genetic sequences. Its scope has specific boundaries that users must understand before applying it to their data.

Within scope

  • Comparative phylogenetics of DNA/RNA sequences (ATCG; U→T conversion required)
  • Alignment-free clustering and dimensionality reduction (PCA, t-SNE, UMAP)
  • Hierarchical dendrogram construction from embedding distances
  • Exploratory analysis of large multi-FASTA datasets
  • Research-grade classification, identification, and relationship inference

Outside scope

  • Protein structure prediction or functional annotation
  • Variant calling or SNP/indel genotyping
  • Quantitative expression analysis (RNA-seq, microarray)
  • Genome assembly or read mapping
  • Clinical genotyping or forensic identification in legal proceedings
  • Any application requiring certified or validated test methods

5. Known technical limitations (as of current release)

The following limitations are known and documented as of the current stable release (v2.1) and beta release (v3.0.1). They do not represent defects but rather the current boundary conditions of the framework.

Encoding
k-mer representation is fixed and non-overlapping

Covary uses a non-overlapping, frequency-independent k-mer encoding. This differs from conventional frequency-based k-mer approaches and may produce different distance rankings when compared against tools that use sliding-window or overlapping k-mer methods.

Input
RNA sequences require manual U→T conversion

Covary processes DNA sequences (A, T, C, G). RNA inputs containing uracil (U) must be manually converted to thymine (T) prior to analysis. Sequences containing characters outside ATCG are filtered out by default when include_N = "no".

Scale
Memory constraints on free-tier compute

Very large datasets (thousands of long sequences) may exceed available RAM on free-tier Google Colab sessions (12–16 GB). v3.0.1 beta introduces improvements for chromosome-level sequences, but production-scale runs may still require a Colab Pro session or local GPU environment.

Taxonomy
Resolution degrades at very high sequence divergence

Covary's embedding-based distance works best when comparing sequences that share a common evolutionary origin. At kingdom-level or domain-level divergence (e.g., mixing bacteria with archaea and eukaryotes in a single run), cluster separation may become unreliable. Uniform, biologically coherent input datasets are strongly recommended.

Reproducibility
t-SNE and UMAP are stochastic — results vary between runs

t-SNE and UMAP projections are non-deterministic unless a fixed random_state is set. Even with a fixed seed, results may vary across different software versions or compute environments. Dendrograms from hierarchical clustering are deterministic given fixed input, but embedding-based spatial positions are not. For reproducible results, always record the random_state value used.

Alignment
Alignment-free does not mean reference-free in all cases

While Covary avoids multiple sequence alignment (MSA), some workflows — particularly identification tasks — implicitly benefit from a consistent sequence start position. The Seed Aligner toolkit is provided to standardize sequence start points. Omitting this pre-processing step on highly variable-start sequences may reduce resolution.

Benchmarking
Comparison against gold-standard tools is limited

Formal head-to-head benchmarking against established phylogenetic tools (e.g., MEGA, IQ-TREE, RAxML, FastTree) is ongoing and limited to the datasets described in the published preprints. Covary should not be assumed to outperform or replace these tools universally — it addresses a different computational niche (alignment-free, large-scale, translation-aware) rather than directly competing on identical tasks.

Platform
Currently delivered as a Google Colab notebook only

Covary v2.1 and v3.0.1 beta are distributed as Google Colab notebooks and are dependent on Google's infrastructure, Python runtime availability, and session limits. There is currently no standalone desktop application, command-line package (pip/conda), or containerized deployment. Session disconnects, runtime resets, or Colab policy changes may interrupt analysis.


6. Limitation of liability

To the fullest extent permitted by applicable law, Marvin De los Santos, ChordexBio, and any contributors to the Covary project shall not be liable for any direct, indirect, incidental, special, consequential, or punitive damages arising from:

  • Use of or inability to use Covary or this website
  • Reliance on any output, result, recommendation, or content generated by Covary
  • Errors, inaccuracies, or omissions in results
  • Loss of data, research time, or resources
  • Any downstream application of Covary outputs in research, clinical, or commercial settings

This limitation applies regardless of whether the liability is based on contract, tort, negligence, strict liability, or any other legal theory, even if the developer has been advised of the possibility of such damages.


7. Website and service limitations

  • This website is provided for informational purposes. Content may be updated, removed, or changed at any time without notice.
  • The interactive simulation on this site is a programmatic demonstration only and does not represent a live execution of Covary's machine learning pipeline.
  • External links (GitHub, Google Colab, preprint servers, ChordexBio) are provided for convenience. We do not control the content, availability, or accuracy of third-party sites.
  • Access to Covary via Google Colab is subject to Google's Terms of Service and infrastructure availability. We cannot guarantee uninterrupted access.
  • The contact form collects name, email, institution, and message for the sole purpose of responding to inquiries. Data is not shared with third parties or used for marketing.

8. Citation and attribution

If you use Covary in research that leads to a publication, presentation, or public dataset, please cite the primary reference:

De los Santos, M. (2025). Covary: A translation-aware framework for alignment-free phylogenetics using machine learning. bioRxiv. https://doi.org/10.1101/2025.11.13.687960

Failure to cite does not void this disclaimer. Attribution is requested as a matter of academic integrity, not as a legal condition of use under the existing license.


9. License requirement

Covary v1.0 – v2.0 is available under a limited non-commercial, for-research-only grant as maybe described in the some published Covary Licenses. Covary v2.1 and all subsequent versions require a valid license for all users regardless of location or affiliation.

Covary (v2.1 and above) is not open-use software. A license (FREE or paid) must be obtained prior to running Covary for any purpose. Three license tiers are available depending on your intended use:

Validation
Validation license

Intended for structured evaluation of Covary's outputs against known datasets or reference methods. Suitable for researchers conducting formal benchmarking, tool comparison studies, or pre-adoption technical assessment within an institution.

Research / Education
Research or education license

For use in academic research, thesis work, graduate studies, classroom instruction, or non-commercial scientific inquiry. This includes participants in the Covary Research Program (CRP). Outputs may be used in publications and academic submissions with proper citation.

Commercial
Commercial license

Required for any use of Covary within a for-profit organization, product development pipeline, consulting engagement, or revenue-generating activity — including internal R&D where outputs inform commercial decisions. Contact us to discuss terms.

How to obtain a license

Licenses are issued through official channels only. To request a license, reach out via either of the following:

Contact form on this site → ChordexBio official channels →

Include your name, institution, intended use, and the license tier you are requesting. License grants are at the sole discretion of the developer and may include usage conditions specific to your project.


10. Changes to this disclaimer

This disclaimer may be updated at any time as Covary evolves. Significant changes will be noted in the release notes at /covary-releases. Continued use of Covary after changes constitutes acceptance of the updated disclaimer.

Last updated: May 2026  ·  Questions? Contact us