Purpose: The rationale of the current study was to develop 6-mercaptopurine (6-MP)-mediated hematological toxicity prediction model for acute lymphoblastic leukemia (ALL) therapeutic management.
Methods: A total of 96 children with ALL undergoing therapy with MCP-841 protocol were screened for all the ten exons of TPMT, exon 2, exon 3 and intron 2 of ITPA using bidirectional sequencing. This dataset was used to construct prediction models of leucopenia grade by constructing classification and regression trees (CART) followed by smart pruning.
Results: The developed CART model indicated TPMT*12 and TPMT*3C as the key determinants of toxicity. TPMT int3, int4 and int7 polymorphisms exert toxicity when co-segregated with one mutated allele of TPMT*12 or TPMT*3C or ITPA exon 3. The developed CART model exhibited 93.6% accuracy in predicting the toxicity. The area under the receiver operating characteristic curve was 0.9649.
Conclusions: TPMT *3C and TPMT*12 are the key determinants of 6-MP-mediated hematological toxicity while other variants of TPMT (int3, int4 and int7) and ITPA ex2 interact synergistically with TPMT*3C or TPMT*12 variant alleles to enhance the toxicity. TPMT and ITPA variants cumulatively are excellent predictors of 6-MP-mediated toxicity.
Keywords: 6-Mercaptopurine; Acute lymphoblastic leukemia; Glutamate carboxypeptidase II; Leucopenia grade; Machine learning; Thiopurine methyl transferase.