Normalization of chemical structure information is a tricky business. One must ascertain scientific intent, often absent vital information. What did the chemist have in mind? Is there more than one interpretation? Potentially making matters worse is that nearly all chemical structure diagrams viewed by humans today were already consumed (i.e., reinterpreted) by computer algorithms one or more times, meaning that the original chemist intent may be long gone. Have you ever used the ‘clean’ function in your drawing package? What did it do to your chemical structure? Did it make your stereocenter interpretation ambiguous or did it interpret defined stereocenters that were previously undefined? Have you ever exported your chemical structure between different file formats and then read each one back in? Did you do this with different software packages? Did your structure still look like what you drew? What stayed the same and what changed?<br/><br/>Sadly, when using standard chemical information desktop tools, one finds that it is increasingly easy to corrupt or lose chemical information in a myriad of ways, often unbeknownst to the user. As humans, we may spend hours drawing and curating chemical structures and, if we are not careful, we can undo that effort in microseconds when we save, export, or copy/paste that chemical structure into a vast array of ‘industry standard’ chemical structure formats that vary in degree of implementation, custom extensions, and common approach between chemical structure aware applications and toolkits. Upon reading this, a myriad of questions may come to mind: How do I know my structure changed? Are there ways to avoid structural data corruption? Who is to blame for any changes? What is my responsibility in this? Is there a forum to improve things?<br/><br/>Using PubChem as a case study, this talk will give an overview of standard pitfalls when handling chemical structure information and best practices to avoid data corruption when communicating this content. In addition, community efforts to improve the plight of chemical structure content everywhere will be provided.

Caution! Normalization can be hazardous to your data health

254th National Meeting (2017)

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.