How Do You Put All Those Data Into A Virtual Tissue?
We don’t.
No more so than nature puts all the necessary biological elements of a living tissue into a fertilized egg. We grow, step by step, a single virtual cell into a virtual tissue.
Importantly, we don’t attempt to model an entire cell; only its essential features as they pertain to a biological understanding of interest, e.g. – a specific pathway, molecule(s), morphology, etc.
The objective is to arrive at a virtual genome which will develop into, support and function as a validated, virtual phenotype.
The result of this effort is a biologically accurate, virtual mechanism which emulates the biological principles necessary to achieve a virtual tissue; one that mirrors the critical behavioral features of its living counterpart.
We don’t extrapolate predicted biological outcomes within any overall mathematical context. As a matter of fact, there is a notable absence of higher level mathematics, e.g. – partial differential equations, in any of the simulated biological building blocks. Our biological representations are generally algebraic in nature, if at all.
Our overall approach is to faithfully emulate biology; to mimic the relevant details within a cell and its immediate environment as they pertain to the specifics of a given study.
We follow the general principle that to accurately understand the result(s) of a process is to first understand the process itself; as it actually occurs, not how we may choose to interpret or describe it. See #BeTheCell.
It is often asked if our approach is similar to “artificial intelligence” (AI) and its commonly used methods of “machine learning” (ML). It is not. In fact, the two are polar opposites; here’s why.
ML is only relevant in the context of a large amount of data; that is, it proceeds from a broad base of information toward a highly focused conclusion of interest within a given data set.
Developmental biology, and our efforts to emulate it, begin at the opposite end of the data spectrum; from a relatively tiny amount of information, i.e. – a (virtual) genome, from which emerges a vastly larger amount of information, e.g. – a (virtual) tissue.
This is not to say that ML cannot help our emulation efforts; it most certainly can. The bioinformatics field and increasingly sophisticated laboratory techniques are producing increasing amounts of relevant data sets. From these it may be possible to arrive at more precise and relevant detail used in our construction of virtual genomes.
We are continually exploring any and all opportunities that may help the advancement and better understanding of our world generally, and life sciences in particular.