Cellular State Image Analysis

Candida albicans is a prevalent, generally commensal member of the human mycobiome carried by most people their entire lives. Simultaneously, C. albicans is a major opportunistic pathogen – in immunocompetent individuals this generally manifests in vulvovaginal (yeast infections) and cutaneous infections while in immunocompromised people, C. albicans can also cause superficial oral infections (thrush) or internal solid organ or bloodstream infections.

The ability to switch between cellular states with distinct metabolic signatures, adhesins, and morphologies enables Candida albicans to invade a variety of host tissues while evading the immune system as well as common treatments. Cell state transition has also been linked to commensal host immune and metabolic regulation. The cellular states we will primarily focus on are yeast-form (round) cells and hyphal (filamentous) cells. Understanding these cell transitions is crucial, but dispersal–the process by which hyphal dominated biofilms generate planktonic yeast-form cells–remains poorly understood. By modeling the regulatory pathways controlling dispersal, we hope to uncover potential targets for treatment to encourage commensalism over pathogenesis; however, quantification of these transitions is challenging by conventional methods.

You will design methods to address this problem — identifying and quantifying cellular state transitions in timelapse microscopy images. On the most basic level, this could be something akin to texture analysis or image classification estimating the quantity of dispersed cells or biofilm per slide in the timelapse. A better solution would identify individual dispersed cells and hyphae enabling analysis of within group heterogeneity such as hyphal length, branching, and twisting or dispersed cell size, clumping, or elongation. An even better solution would additionally track dispersed cells, enabling analysis of post-dispersal growth.

Submissions are evaluated on accuracy and richness of information:

Minimal model:
Initiation of dispersal:
0-2 points are awarded for accuracy in identifying the slide in which dispersal begins

Biofilm growth:
0-2 points are awarded for accuracy in creating a growth curve of the biofilm over time

Dispersed cell quantification:
0-2 points are awarded for accuracy in quantifying the number of dispersed cells per slide

Icing on the cake:
Dispersed cell characterization:
0-.5 points are awarded for identification of individual dispersed cells with an additional 0-.5 points for their characterization (cell size, clumping, or elongation, etc)

Biofilm characterization:
0-1 points are awarded for the characterization of yeast form and hyphal cells within biofilms (distribution of yeast-form associated cells along hyphae, length, branching, and twistiness of hyphae, etc)

Cell tracking:
0-1 points are awarded for tracking individual dispersed or biofilm associated cells throughout the timelapse, enabling quantification of their growth and development