Data selection choices influence the inferred movement patterns of Plasmodium sporozoites in skin

Motility of Plasmodium sporozoites (SPZs) in the skin is a key determinant of successful host infection. Earlier studies have described rapid movement of both murine and human SPZs in skin following syringe inoculation. It is typical to classify SPZ trajectories into ``motile'' and ``immotile'' and restrict the analysis of movement patterns to motile SPZs. Because criteria to define motile SPZs are dependent on the study and are often qualitative, it remains unclear if sub-selection of motile tracks introduces biases in characterization of SPZ movement in vivo. We processed imaging data (22 movies) from a recent study of movement of P. falciparum (Pf) and P. yoelii (Py) SPZ in skin. We proposed a novel metric -- maximal spatial spread (MSS or S) --- that is the maximum Euclidean distance between any two recorded positions in a trajectory. We used MSS to classify SPZ trajectories as immotile (S<Sthreshold) or motile (S>Sthreshold) for a given threshold value Sthreshold. Larger Sthreshold values naturally resulted in a smaller fraction of tracks classified as motile, and subsequently, in an increased overall displacement, instantaneous and mean speeds, decreased mean turning angle, and higher initial slopes of the mean squared displacement (MSD) curves. We found that at intermediate values of Sthreshold Pf SPZs had a lower average speed than Py SPZs suggesting that host environment may impact SPZ movement. Both species exhibited a small but statistically significant decline in average speed with time after inoculation but this was also dependent on the Sthreshold value. Our analysis of MSD curves and turning angle distributions suggests that both Pf and Py SPZs undergo correlated random walks -- a type of Brownian walk with short-term superdiffusive displacement. By using a novel methodology of hidden Markov models (moveHMM package in R) we found that SPZ movement is best described by three movement states; however, none of these states corresponded to previously described circling gliding. Taking together, our results suggest that inference of SPZ movement patterns depends on the criteria used to define tracks as motile or immotile. Standardized preprocessing criteria are therefore important when comparing motility across Plasmodium species, experimental time points, or laboratories. Analysis of turning angle distributions and application of hidden Markov models provided additional metrics to quantify distinct modes of SPZ movement in vivo.