RESUMO
Nuclear maturation is an essential process in which oocytes acquire the competence to develop further. However, the time required for nuclear maturation during IVM varies among oocytes. Therefore, predicting nuclear maturation speed (NMS) could help identify the optimal timing for IVF and maximize the developmental competence of each oocyte. This study aimed to establish machine learning-based prediction models for NMS using non-invasive indicators during the individual IVM of Japanese Black (JB) beef heifer oocytes. We collected ovaries from abattoirs and aspirated cumulus-oocyte complexes (COCs) from follicles with diameters ranging between 2 and 8 mm. The COCs were matured individually for 18 h, and photographs of each COC were taken at the beginning and every 3 h from 12 h to the end of maturation. After IVM culture, we denuded COCs and stained oocytes to confirm the progression of meiosis. Only oocytes that reached the metaphase II (MII) stage were considered to have a fast NMS. Morphological features, including COC area, cumulus expansion ratio, expansion rate per hour, and expansion pattern, were extracted from the recorded photos and applied to develop prediction models for NMS using machine learning algorithms. The MII rates of oocytes with fast- and slow-predicted NMS differed when the decision tree (DT) and random forest (RF) models were employed (P < 0.05). To evaluate the relationship between predicted NMS by DT and RF models and fertilization dynamics during individual IVF, sperm penetration and pronuclear formation were evaluated at 3, 6, 9, and 12 h after IVF start, following 24 h of IVM. The association between predicted NMS and embryo development was investigated by performing IVC for seven days using microwell culture dishes after 24 h of IVM and 6 h of IVF. Predicted NMS did not show a significant association with fertilization dynamics. However, oocytes with fast-predicted NMS by the RF model exhibited a tendency towards a higher cleavage rate 48 h after IVF start (P = 0.08); no other relationship was found between predicted NMS and embryo development. These findings demonstrate the feasibility of using non-invasive indicators during IVM to develop prediction models for NMS of JB beef heifer oocytes. Although the effect of predicted NMS on embryo development remains unclear, customized treatments based on NMS predictions have the potential to improve the efficiency of in vitro embryo production following individual IVM culture.
