FAQs

Is Discovery based on topic models and LDA (Latent Dirichlet Allocation)?

Yes, the core method behind Discovery is LDA. However, obtaining robust and reproducible results from LDA is notoriously hard. This is partly due to the probabilistic nature of the method. This is a critical limitation in biomedical applications, where reproducibility of findings is key. The added value of Discovery lies in the fact that it runs an extensive set of experiments from which the optimal number of cliques, as well as the most robust solution, are identified using a validated criterion.

 

Is Discovery just a clustering algorithm?

No, microbial cliques computed by Discovery are not clusters. In fact, microbial cliques are probability distributions over the taxa. The properties of such a distribution carry a whole lot of information about the biological process or mechanism encoded by the clique.

 

I have counts instead of relative abundances, can I use Discovery?

Yes, you can use either counts (either already scaled or not scaled) or relative abundance data.

 

I learned the set of microbial cliques in my data. How do I find the meaning of each clique? In other words, to which processes or functionalities are the cliques associated?

Our Demo does not provide this feature yet. However, microbial cliques group the information in a way such that a person with domain knowledge can easily associate a meaning to each clique. When we work with our clients and partners, we provide this feature by looking at specific functional modules or pathways that are enriched in each clique.

 

Is there a scientific publication associated with Minutia Discovery?

At the moment there is no publication associated with Discovery. We are reserving the method for commercial use as we continue our research and development aimed at adding more functionalities, such as handling time-series data and metadata. However, please read our white paper for more detailed information about how Discovery works. And we are always happy to answer more specific questions if you contact us.

 

Is Minutia.ai using my data to train its own large models?

No, your data is not saved anywhere and the results are displayed to you only. Once you close the webpage, the memory is flushed and there won’t be any trace left of either the input data or the results. We are also not saving any intermediate results or model parameters. The only data we save is the number of accesses to the tool via the web page and how much computing power is used by each run.

 

Can I run Discovery with other types of Omics data?

Although the entire tutorial and explanations make reference to microbial abundance data (taxonomical abundances), Discovery can work with any type of omics data that follows the same input format (functionality data, metabolomics, etc.). However, if you are planning to use Discovery for data other than microbial abundances, we suggest you consult with us first.

​

I have time-series data, can I use Discovery?

At the moment, Discovery cannot handle time-series data as we are still working on it. Once the functionality is ready, it will be extremely informative to see how the same clique can evolve over time (in terms of its members and their within-clique importance). For example, a clique responsible for metabolizing carbohydrates will have its most important members shifting from species able to break down breast milk to species able to process more complex carbohydrates as a baby goes through its first years of life. The same reasoning would apply to a clique associated with the progression of a disease or any other time-dependent process. Although we are not ready to launch this functionality as part of an automated tool, we do perform time-dependent analyses with our collaborators and for our clients.

 

I have metadata associated with my taxonomic abundance data, can I link it to the microbial cliques?

Similarly to time-dependent cliques, we are still working on making cliques dependent on specific metadata. Once the functionality is ready, it will be possible to observe how the same clique changes (in terms of its members and their within-clique importance) depending on diet, lifestyle or geographic location. Although we are not ready to launch this functionality as part of an automated tool, we do perform metadata-dependent analyses with our collaborators and for our clients.

 

Why is Discovery demo limited to 100 samples?

Discovery requires a fair amount of computational power in order to identify a robust and reproducible solution in your data, so this is a way for us to control cost while welcoming anyone to explore our demo.

 

Can I download the results of my analysis?

This demo displays results on screen but does not allow you to download results or reports. Please get in touch with us if you feel Discovery could be of help to you or to your organization.