248
P. Volarath et al.
export. This workflow exports the chemical/toxicity/regulatory data of the selected
CERES chemical in a portable document format (PDF) document (Fig. 12.5b). The
structure data (SD) can also be exported as an SD file that can be processed by external cheminformatics tools. Additional informatics workflows such as read-across
and weight-of-evidence calculations are under consideration for development.
12.3 Current Challenges and Future Directions
Although the computational approaches described above exhibit great potential for
improving the efficiency of the safety assessment process and expanding our understanding of potential toxicity of food ingredients, these methods have limitations that
need to be recognized and carefully examined.
12.3.1 QSAR Modeling
QSAR modeling has many challenges [60]. One challenge related to QSAR model
development for food ingredient safety evaluation is the limitation of current software
to handle inorganic and metal-containing compounds, salts, and mixtures. An experiment was conducted on CFSAN’s public food ingredient inventories (Table 12.2)
[61–66]. The chemical records from these inventories were screened for inorganic,
organometallic, metal complexes, metals, and polymers. From the 19,265 total
records, 10,884 records had definable structures that were further reviewed by OFAS
scientists. The structures of these 10,884 records were then programmatically processed to remove small inorganic fragments for those structures that contained multiple chemical components, neutralize charged molecules, generate 3D structures,
and flag duplicates. This analysis removed 8609 structures and yielded 2275 final
food ingredients whose structures could be used for QSAR analysis. This workflow
is typically performed to systematically remove chemicals that cannot be efficiently
processed by cheminformatics software. As shown in this example, out of the starting
19,265 food ingredient records, only 2275 (~12%) had structures that could be used in
the analysis. This example indicates the need to explore and identify issues within the
excluded chemical records and structures. At the same time, it illustrates current challenges in the softwares’ abilities to handle polymers and inorganic/metal-containing
compounds, which also need to be further explored. In the future, as more reliable
data on polymers and inorganic/metal-containing compounds become available, it is
expected that the structure handling capabilities in the software will improve. The
software enhancements, together with the knowledge gained from examining the
excluded chemical records, will enable more of the food ingredients that currently
fall under these categories to be processed and used to improve QSAR models for
the safety assessment of food ingredients.
P. Volarath et al.
export. This workflow exports the chemical/toxicity/regulatory data of the selected
CERES chemical in a portable document format (PDF) document (Fig. 12.5b). The
structure data (SD) can also be exported as an SD file that can be processed by external cheminformatics tools. Additional informatics workflows such as read-across
and weight-of-evidence calculations are under consideration for development.
12.3 Current Challenges and Future Directions
Although the computational approaches described above exhibit great potential for
improving the efficiency of the safety assessment process and expanding our understanding of potential toxicity of food ingredients, these methods have limitations that
need to be recognized and carefully examined.
12.3.1 QSAR Modeling
QSAR modeling has many challenges [60]. One challenge related to QSAR model
development for food ingredient safety evaluation is the limitation of current software
to handle inorganic and metal-containing compounds, salts, and mixtures. An experiment was conducted on CFSAN’s public food ingredient inventories (Table 12.2)
[61–66]. The chemical records from these inventories were screened for inorganic,
organometallic, metal complexes, metals, and polymers. From the 19,265 total
records, 10,884 records had definable structures that were further reviewed by OFAS
scientists. The structures of these 10,884 records were then programmatically processed to remove small inorganic fragments for those structures that contained multiple chemical components, neutralize charged molecules, generate 3D structures,
and flag duplicates. This analysis removed 8609 structures and yielded 2275 final
food ingredients whose structures could be used for QSAR analysis. This workflow
is typically performed to systematically remove chemicals that cannot be efficiently
processed by cheminformatics software. As shown in this example, out of the starting
19,265 food ingredient records, only 2275 (~12%) had structures that could be used in
the analysis. This example indicates the need to explore and identify issues within the
excluded chemical records and structures. At the same time, it illustrates current challenges in the softwares’ abilities to handle polymers and inorganic/metal-containing
compounds, which also need to be further explored. In the future, as more reliable
data on polymers and inorganic/metal-containing compounds become available, it is
expected that the structure handling capabilities in the software will improve. The
software enhancements, together with the knowledge gained from examining the
excluded chemical records, will enable more of the food ingredients that currently
fall under these categories to be processed and used to improve QSAR models for
the safety assessment of food ingredients.