Cheminfostream

Pharma R&D Data Management using an Enterprise Electronic Lab Notebook Approach

A very wide variety of information is generated during the process of Pharma R&D; this information needs to be captured and stored for intellectual property (IP) purposes, but more importantly, needs to be utilised to drive the future scientific research efforts. Capturing the information generated by research staff by using electronic versions of the paper laboratory notebook is a rapidly evolving application area.

The types of data to be handled, and the means by which the data is to be archived, reported, validated and utilised are also changing very rapidly. In his presentation Robert Scoffin of CambridgeSoft gives an overview of the Electronic Laboratory Notebook (ELN) application area to highlight the advantages of an ELN over a paper-based system.

Scoffin points out that the pharma and biotech industry need ELN solutions that satisfy long-term data archival requirements, improve productivity but maintain controls on costs.  Although an estimated 90% of systems currently followed a mixed lab notebook model of paper and electronic, an increased use of electronic-only systems is underway.

ELNs enable knowledge obtained from previous research projects with molecules to be captured and aid decision-making in new projects.  Creation of a future-proofed knowledge store provides a valuable long-term asset for the organisation from which they can be seeing returns 10 to 20 years on.  ELNs must also cope with increasingly complex regulatory requirements and intellectual property must be captured and protectable from a legal point of view. 

Integrating electronic lab notebooks and knowledge management systems to accelerate discovery

We now have the talk "Integrating electronic lab notebooks and knowledge management systems to accelerate discovery" delivered by Lorie Karnath, Managing Director International and Jeff Spitzner, Chief Scientific Officer, Rescentris available on the Cheminformatics and Modelling website site.  The problem of many different incompatibilities between data formats provides a major challenge to the pharmaceutical industry.  At a time when knowledge integration and sharing, collaboration and business agility are becoming increasingly important, proprietary formats, incompatible systems and sheer data volume and complexity provide significant barriers to progress. In this Virtual Seminar we hear about the progress being made in the Life Science and bioinformatics areas so as to enable XML-based interoperability within the context of R&D and electronic laboratory notebook systems.  The effort includes a new $6 million public-private partnership involving Ohio State University and Rescentris to enable greater progress in biomedical research.  When I questioned Jeff Spitzner on how the significant barriers to integration over a large number of equipment and IT vendors in the bio, chemistry and medical information spaces could be overcome, he admitted that this could not be achieved in a week or even a year, but that the pharmaceutical industry should supply pressure to drive the market in this direction.

Quality Control in QSAR Model Development

Quantitative Structure Property Relationship (QSPR) modelling finds growing applications in chemical data mining and combinatorial library design. The Cheminformatics & Modelling website presentation of Alex Tropsha on Quality Control in QSAR Model Development emphasises the importance of rigorous validation as a crucial component of QSPR model development. He presents a set of simple guidelines for developing validated and predictive QSPR models.

Alex Tropsha explains in his talk how to build Quantitative Structure-Activity Relationship (QSAR) models that can be validated and used for external database mining and drug discovery. He points out that the quality of models is often typically based on internal training sets alone, but that this approach does not necessarily generate good predictive models.  Rather, regardless of which combinations of descriptor sets and correlation methods (linear and non-linear) are used, a quality assessment approach should be applied to the selection of really useful predictive models for drug discovery.

As an example, he shows how a published CoMFA model for PK3A4 inhibitors (ecdysteriod dataset), has a good fit for its training set, but fails the test of being a good predictive model when applied to an external test set.

He discusses several validation strategies including (1) randomization of the modelled property, also called Y-scrambling, (2) external validation using rational division of a dataset into training and test sets, and (3) identification of the model applicability domain in the chemical space to flag molecules for which predictions may be unreliable.