DeepPatterns wins BioTech track in Open Innovations Startup Tour
Our project “DP4SAR — the innovative tool for discovering chemical structure patterns and structure-activity relations for directed drug design” awarded first prize in “BioTech in Medicine and Agriculture” section.
The final round of Saint Petersburg track of “Open Innovations Startup Tour” — the biggest contest for innovative startups in Russia and CIS — has been held on Aplril, 19 at LenPoligrafMash technopark.
According to organizers the contest became replete in sense of applicants count (more than 140 startups attended the event) as well as in competition intensity. Saint Petersburg vice-governor Evegeni Elin recognized the Startup-tour as “very important for the city”.
Our company attended the “BioTech in Medicine and Agriculture” track with “DP4SAR — the innovative tool for discovering chemical structure patterns and structure-activity relations
for directed drug design” project.
This solution addresses actual fundamental scientific and practical problem — discovering Structure — Activity relationship for chemical compounds.
The presentation made by Andrei Konchakov, DeepPatterns CEO, explained the reasons for insufficient effectiveness of legacy mathematical methods applied to SAR problem from both formal model accuracy criteria and model interpretability views. It was emphasized that uninterpretable models can’t be used for directed drug design.
It was also noted that logical rules discovery algorithms are the best fit to systematic complexity of experimental data. At the same time implementations of well-known algorithms are limited in analysis depth because finding the “strong” logical rules is extreme combinatorial problem and accompanied by so-called combinatorial explosion.
DP4SAR project implements an innovative method of finding logical rules. Algorithms based on SRMD method utilizing the resonance phenomenon in data invoked by special procedures are able to reveal so-called “strong” (most complete with given accuracy) rules in heterogeneous multidimensional data. These rules reflect complex systemic relations in data that are impossible to discover by traditional algorithms like decision trees and limited search.
SRMD method significantly surpasses known data mining methods aimed at finding logical rules in data in terms of accuracy and completeness. It is also very important that the method allows high degree of parallelism in computations that determines its success in BigData area.
All these advantages of SRMD define success of DeepPatterns in solving actual data science problems in a whole and in bioinformatics in particular.
The presentation made at StartupTour contained description of DP4SAR prototypes and examples of solved problems that shown advantages of DP4SAR — higher accuracy and performance as well as interpretability of results allowing directed drug design approach.
DP4SAR won the first place in “BioTech in Medicine and Agriculture” track by vote of expert committee. DeepPatterms representatives were invited to Startup Village — the biggest startup conference in Russia and CIS that will be held on 29-30 of May, 2019 at Skolkovo innovation center.