Prediction of serum biochemistry of pigs at pre- and post-weaning stages of growth using regression tree methods

This study was aimed at evaluating the effects of breed, sex and their interactions on aspects of haematology, serum biochemistry and morphometry of healthy pigs at pre- and post-weaning stages of growth (4 and 12 weeks respectively). It was also aimed at predicting the serum biochemical parameters (SBP) that are biomarkers for growth using regression tree method (RTM). Seventy-one crossbred (71 CB) pigs (Landrace × Local pigs) and 21 Local (L) pigs were reared from birth to adult stage, with blood samples and morphometric parameters (MP) taken at pre- and post-weaning stages. The CB pigs had higher values for aspartate transaminase (AST) (serum glutamic oxaloacetic transaminase, SGOT) (P < 0.05) than local ones, while the female animals had higher values (P < 0.05) than the males at pre-weaning stage of growth. At post-weaning stage, the CB breeds also had higher AST (SGOT) values (P < 0.05) than the L breeds, while among the sexes, the males had higher values for WBC, L, mean corpuscular haemoglobin and mean corpuscular haemoglobin concentration (P < 0.05) than the female pigs, but the females were higher for N and AST SGOT (P < 0.05) than the males respectively. However, the prediction of SBP using Exhaustive CHAID algorithm of RTM returned 100% prediction in terms of a single terminal node from all the morphometric parameters at both stages of growth. The terminal nodes of the RTM produced mean values of 34.57 UI/L for alkaline phosphate, 29.25 UI/L for AST and 60.27 UI/L for alanine transaminase (ALT) respectively. This implies that the prediction of ALKPOS, AST (SGOT) and ALT (serum glutamic pyruvic transaminase) using the 5 morphometric traits measured is not significant as no single node was linked to them. Although the node values reported for them are within the range of 0–40 IU/L for healthy pigs, the values for ALT are above this range. Therefore, individual contributions from the morphometric traits could not predict any SBP as the model returned zero node.