If you have calculations or a theoretical basis for your product, a virtual prototype is probably still okay for a small initial investment or grant. Beyond this, and now especially in light of everything happening with Theranos, scientific due diligence will play a big role in moving forward.
The single biggest learning point Genesis DNA(GFSA 2105 cohort) had was related to their development timeline: do not plan your development timeline as you would an academic research project. There really isn't time or money to optimize everything along the way, to establish a control experiment for every point of investigation, or to run experiments in serial. Identify the single biggest de-risking milestone and be off to the races. Instead, let there be errors, let there be bugs, let it not work half the time...just build. This can be tough to shift into this mindset if you're coming straight from academia.