Surrogate data – a secure way to share corporate data

Abstract

The privacy of chemical structure is of paramount importance for the industrial sector, in particular for the pharmaceutical industry. At the same time, companies handle large amounts of physico-chemical and biological data that could be shared in order to improve our molecular understanding of pharmacokinetic and toxicological properties, which could lead to improved predictivity and shorten the development time for drugs, in particular in the early phases of drug discovery. The current study provides some theoretical limits on the information required to produce reverse engineering of molecules from generated descriptors and demonstrates that the information content of molecules can be as low as less than one bit per atom. Thus theoretically just one descriptor can be used to completely disclose the molecular structure. Instead of sharing descriptors, we propose to share surrogate data. The sharing of surrogate data is nothing else but sharing of reliably predicted molecules. The use of surrogate data can provide the same information as the original set. We consider the practical application of this idea to predict lipophilicity of chemical compounds and we demonstrate that surrogate and real (original) data provides similar prediction ability. Thus, our proposed strategy makes it possible not only to share descriptors, but also complete collections of surrogate molecules without the danger of disclosing the underlying molecular structures.

Building attention and edge message passing neural networks for bioactivity and physical–chemical property prediction

“…This type of method also presents an approach for the secure sharing of chemical data, i.e. it is possible to disseminate trained models for activity predictions without the risk of reverse-engineering IP-sensitive structural information [31][32][33].…”

Descriptor collision and confusion: Toward the design of descriptors to mask chemical structures

“…Our attempts, focused strictly on therapeutically relevant biological targets (e.g., serine proteases, nuclear hormone receptors, class A G-protein coupled receptors, etc. ), complement the effort of developing a safe exchange system for physico-chemical properties such as LogP [8], where actual disclosure of chemical structures may not be required.…”

“…It is generally believed that one can reverse engineer chemical structures from descriptors [4]. Although such claims were not documented in peer-reviewed literature, one can imagine the use of a problem-solving algorithm such as a genetic algorithm [5] in conjunction with a large database, e.g., ChemNavigator's iResearch Libraryä [6] or Beilstein [7], where appropriate descriptors are computed; with this procedure, one could conceivably converge to a structure that -given the descriptor space -is remarkably similar, or perhaps identical, to the ''target structure'' (i.e., match or come close to the unknown structure's fingerprint) with relative ease [8]. Such claims are now documented in respect to, e.g., the Signature descriptor system [9] and lead to the overall conclusion that it is not safe to release chemical information given the risk that structures might be reverse-engineered, in particular since the party releasing such information desires to keep its chemical structures secret.…”

BIGCHEM: Challenges and Opportunities for Big Data Analysis in Chemistry

“…Even large pharma companies can accumulate only limited amounts of relevant property information. As it was mentioned before, sharing data collected by different organizations offers the opportunity to develop computational models on a much broader data basis, thereby increasing model robustness, accuracy and coverage of chemical space ,. The development of approaches to predict ADME/T properties in a collaborative manner is becoming a part of future pharma R&D strategies.…”