Will HILIC separate your molecules?

The prediction tool is a Java Applet computer program that runs within your web browser. It uses QSRR PLS prediction model information developed by Merck Millipore SeQuant using the multivariate modeling software UmBio Evince.

All calculations and predictions of retention times are performed locally on your computer, and all information on chemical structures are kept within the prediction program and is discarded when you clear the list of compounds or close the web browser window.

• Start program
  Click on the link below to open the page with the HILIC Retention Prediction program. The prediction tool will automatically start within your web browser and download the most recent modeling information from Merck Millipore SeQuant and UmBio. You must have Java from Sun Microsystems  installed on your computer and accept the digital signature from Merck Millipore SeQuant.
   
• Select structure
  You can input structures either as SMILES text strings or import them as SDF molecular structure files. There are also some structures built into the program that you can access through the drop-down menu at the top. SMILES and SDF information can be generated by many commercial chemical drawing programs and can also be found in the PubChem Compound database  maintained by the National Center for Biotechnology Information in the USA. Search your compound by name in the database, paste the SMILES into the field, and finally press 'Import into table' to import structure information.
   
• Predict retention
  You can select between two different retention models, one at pH 6.7 and one at pH 5.6, using the model drop-down field (but there are only minor differences between the two). Press 'Predict' to calculate retention for your structures or press 'Clear' to empty the table and make room for new structures. For compounds with predicted negative retention the program will show 'No retention' or 'Possible retention', depending on the size of the negative value.

How does the tool predict retention?

QSRR (quantitative structure-retention relation) prediction models are based on retention data of 40 different hydrophilic compounds, many of which are of plant metabolite origin. Multivariate PLS (partial least square) models were developed with UmBio Evince software. Retention data were obtained on ZIC®-HILIC columns (100 x 4.6 mm, 5 µm, 200 Å) using an eluent containing 70% acetonitrile and 30% 100mM ammonium acetate buffer, pH 6.7 or pH 5.6 (two different models).

The models show good fit to the data (R2=0.86-0.87) and adequate prediction ability (Q2=0.64-0.67).