Reproducibility

every computational method has assumptions what are yours? and why should i believe they hold true for my biological question? our core assumption is that genes whose expression is consistently correlated across large numbers of independent samples and studies are functionally related this is the foundational principle of co expression network biology, and it is well supported by decades of evidence showing that co expressed genes tend to share biological functions, participate in the same pathways, and be co regulated by common upstream mechanisms beyond that, our networks assume that building tissue specific and condition specific models instead of one universal network better reflects biological reality, since gene regulatory relationships are known to vary across tissues and disease contexts what we deliberately avoid assuming is equally important we do not assume that previously published interactions are correct (we don't use literature curation) we do not assume that all gene relationships are linear (which is why we also offer ai networks that capture non linear correlations) we do not assume that any single dataset is reliable on its own every interaction in the platform must be supported by statistically significant signal across integrated data from multiple independent studies whether these assumptions hold for your specific biological question is something the platform helps you evaluate directly the functional annotation card lets you inspect which conditions and datasets support each interaction, the conditions expression chart lets you check gene behavior in your disease of interest, and the dual network comparison lets you test whether an interaction is context dependent or generalizable in other words, the platform doesn't ask you to completely trust the assumptions but gives you the tools to test them against your own research context if i run the same query twice, will i get identical results? yes the networks are precomputed from the underlying data, so querying the same genes with the same parameters in the same network will produce identical results every time there is no stochastic element in the query process the platform is retrieving a defined set of interactions from an established network model, not recalculating correlations on the moment your results are fully reproducible within a given data version the only scenario where results would change is if the underlying network has been updated with new data or improved curation between your two queries in that case, the difference reflects an improved data foundation, not inconsistency how do i know the parameters you use? there are so many risky decisions you need to make and disclose when performing these analyses? how can i trust your science? any computational platform involves a chain of methodological decisions normalization choices, correlation methods, thresholds, filtering criteria, etc our approach is to keep the science open and the engineering proprietary the methodological foundations of the platform, or how networks are built, what statistical frameworks underpin the correlations, how quality control is performed, and how the models have been validated are documented and accessible through the background analyses and approaches docid\ fvmeheyw30j fuvdmqcri sections we encourage you to review them and challenge them every methodological decision in the pipeline has been validated by our r\&d team against established biological benchmarks what remains proprietary is the engineering and backend infrastructure the platform architecture, optimization, and scaling systems that make it possible to build and serve these networks at the scale we do this is our intellectual property, but it is distinct from the scientific methodology, which we believe should be open for review if you have specific questions about a particular methodological choice why a certain threshold was used, how a normalization step was implemented, or how a validation was performed, our r\&d team will be happy to discuss them directly can you provide the exact version of every software package and database used in my analysis? we maintain version control across our data and network builds each network version is documented, including the data sources and database versions that contributed to it this information is available through the platform for the specific software packages and internal pipeline tools used in the backend, these fall under our proprietary engineering infrastructure and are not disclosed at the package level however, the methodological approach including the statistical methods, normalization procedures, and validation frameworks is documented and transparent if you need version level detail for a specific publication or regulatory context, reach out to our team and we can discuss what can be provided on a case by case basis how do i include your analyses in my methods section? our methodology is documented on in the background analyses and approaches docid\ fvmeheyw30j fuvdmqcri section, which provides the information you need to describe the analytical approach in your own methods section, including the network construction framework, the data sources, and the statistical analyses to cite mavatar discovery, read more under cite mavatar discovery docid\ q9i dwmoz2zfligorig8g here are some examples the graph was generated based on the \[tissue, context] (version 1 0) network using mavatar discovery v1 1 0 (2025 mavatar, https //discovery mavatar com https //discovery mavatar com ) the graph was generated based on the combined \[1st tissue, context] (version 1 0) and \[2nd tissue, context] (version 1 0) networks using mavatar discovery v1 1 0 (2025 mavatar, https //discovery mavatar com https //discovery mavatar com ) the analyses were done based on the \[tissue, context] (version 1 0) network using mavatar discovery v1 1 0 (2025 mavatar, https //discovery mavatar com https //discovery mavatar com ) for more specific citation guidance, feel free to contact us directly we're happy to help you draft the appropriate methods description for your manuscript and ensure it meets the reporting standards of your target journal how do i cite mavatar discovery? to cite mavatar discovery, read more under cite mavatar discovery docid\ q9i dwmoz2zfligorig8g here are some examples the graph was generated based on the \[tissue, context] (version 1 0) network using mavatar discovery v1 1 0 (2025 mavatar, https //discovery mavatar com https //discovery mavatar com ) the graph was generated based on the combined \[1st tissue, context] (version 1 0) and \[2nd tissue, context] (version 1 0) networks using mavatar discovery v1 1 0 (2025 mavatar, https //discovery mavatar com https //discovery mavatar com ) the analyses were done based on the \[tissue, context] (version 1 0) network using mavatar discovery v1 1 0 (2025 mavatar, https //discovery mavatar com https //discovery mavatar com ) what happens when you update your networks or database? can i still access my old analyses? network updates reflect improvements in data coverage, curation, and methodology they are a feature of a growing platform, not a disruption to your work each network version is documented, and previous versions are retained within the platform, so you can still access and query older network builds if needed this means that if you ran an analysis and published results based on a specific network version, you can return to that same version for reproducibility when a new version becomes available, you also have the option to rerun your analysis against the updated networks to see how expanded data or improved curation affects your results if you need specific guidance on versioning for a publication or regulatory context, our team is happy to help how do i explain to reviewers that i can't share the code because it is proprietary? can a reviewer access my analyses without subscription? this is a common concern when using any commercial platform in academic research, and it's not unique to mavatar discovery the same applies to tools like ingenuity pathway analysis (ipa), metacore, or any licensed bioinformatics software in your manuscript, you describe the methodology and parameters used, cite the platform, and reference the documented methods available under background analyses and approaches docid\ fvmeheyw30j fuvdmqcri you are not expected to share proprietary source code any more than you would be expected to share the internal codebase of ipa or any other commercial tool what matters for reproducibility is that the analytical approach is transparent, the parameters are reported, and the results are traceable, all of which the platform supports regarding reviewer access, contact our team we can discuss options for providing reviewers with temporary access to the platform so they can evaluate your analyses directly we understand that peer review requires scrutiny, and we are prepared to support that process on a case by case basis