A new Bayesian network-based approach to the analysis of sperm motility: application in the study of tench (Tinca tinca) semen.

In this study a Bayesian network (BN) has been built for the study of the objective motility of Tinca tinca spermatozoa (spz). Semen from eight 2-year-old sexually mature male tenchs was obtained and motility analyses were performed at 6-17, 23-34 and 40-51 s after activation, using computer-assisted sperm analysis (CASA) software. Motility parameters rendered by CASA were treated with a two-step cluster analysis. Three well-defined sperm subpopulations were identified, varying the proportion of spermatozoa contained in each cluster with time and male. Cluster, cinematic and time variables were used to build the BN to study the probabilistic relationships among variables and how each variable influenced the final sperm classification into one of three predefined clusters. Both network structure and conditional probabilities were calculated based on the collected data set. Results shown that almost all the variables were directly or indirectly related to each other. By doing probabilistic inference we observed that the cluster distribution corresponded to the definition provided by the cluster analysis. Also, velocity and time variables determined the cluster to which each spermatozoon belonged with a high degree of accuracy. Thus, BNs can be applied in the study of sperm motility. The construction of a BN that include fertility data opens a new way to try to clarify the roles of motility and other sperm quality indicators in fertilization.