|
1B+C Abiotic Transformations + Biodegradation (MOP/27) Using ready biodegradation data to validate BIOWIN. Boethling, R.S.1, Lynch, D.G.1, Thom, G.C.1, Tunkel, J.L.2, 1 U.S. Environmental Protection Agency, Washington, DC, USA2 Syracuse Research Corporation, Syracuse, NY, USA ABSTRACT- Whether a given chemical substance is readily or not readily biodegradable (RB or NRB) is an important piece of information in risk screening for both new (premanufacture) and existing chemicals. Despite the relatively low cost of OECD ready tests, data are often unavailable and a need exists for computer methods that can estimate this endpoint reliably. In previous work we showed that predictive accuracy for 305 Premanufacture Notice (PMN) substances was in the 80+ % range for 4 of 6 Biowin models (MITI linear; MITI nonlinear; Survey ultimate; Survey primary). Here we focus on Biowin3 (Survey ultimate) and Biowin5 (MITI linear) using an expanded database of 374 PMN substances, plus a carefully checked file of 398 existing chemicals with MITI (OECD 301C) test data. Biowin3 was the best performer with PMN data: 86% correctly classified for both 96 RB and 278 NRB chemicals. Using the evaluated MITI data, Biowin3 accuracy was slightly lower (80%) but comparable to that reported for the Biowin5 training set (Tunkel et al. 2000). For the PMN data the percentage of Biowin3 RB predictions that were correct (i.e. predictive value positive; PVP) was lower (69%), but false positives were substantially reduced if a two-model approach was applied in which Biowin3 RB predictions were subjected to confirmation by Biowin5 (PVP = 79%). Overall accuracy was 88% and all chemicals were estimated. PMNs with (i) multiple ready tests or (ii) both inherent and ready biodegradation tests were also studied. Limited data were available for the comparisons, but the results at once enhance the credibility of ready tests and question the value added by inherent tests. Key words: BIOWIN, Ready biodegradation, Premanufacture Notice, MITI |
