Review: Forensic Analysis of the Microbiome of Phones and Shoes

Emily C. Lennert





microbiome, microbe, bacteria, genus, phone, shoe, floor, transfer, DNA

Article Reviewed

  1. Lax, S.; Hampton-Marcell, J. T.; Gibbons, S. M.; Colares, G. B.; Smith, D.; Eisen, J. A.; Gilbert, J. A. Forensic analysis of the microbiome of phones and shoes. Microbiome. 2015, 3(21).


The opinions expressed in this review are an interpretation of the research presented in the article. These opinions are those of the summation author and do not necessarily represent the position of the University of Central Florida or of the authors of the original article.


Existing research has shown that human interaction with surfaces, as well as surface to surface interaction, leads to the transference of microbes, i.e. bacteria. Research has indicated that humans transfer their microbiome, i.e. microbial community, from their skin to the surface of their phones. Additionally, the microbiome of humans has been shown to provide some discriminating power; human microbial signatures have been used in some cases to match individuals to objects that the individual had recently interacted with, such as a computer keyboard. This indicates the potential to use the microbiome as a forensic tool.

The purpose of this study was to determine the impact that the individual and surface type, i.e. phone, shoe, and floor, have on the microbiome observed in a sample. Initially, two participants collected samples hourly, over the course of 12 hours, for 2 consecutive days from their shoes, the floor, and their phones. Samples from shoes were taken in four areas, the right shoe tip, left shoe tip, right shoe heel, and left shoe heel. Floor samples were collected immediately adjacent to the position that the participant stood when shoe samples were collected. Phone samples were taken from the front and back of the phone. Samples were collected with pre-moistened sterile cotton swabs, moistened with a 0.15 M saline solution, and immediately placed in a freezer at -20 ˚C or placed on dry ice until transferred to the freezer. The microbiome content was then determined by DNA analysis. Total DNA was extracted from each swab, quantified, and processed using the Earth Microbiome Project barcoded primer set followed by additional processing and sequencing to determine the microbial content of the sample.

The microbiome was determined to be dependent on both the individual and the sampling surface, with sampling surface providing more distinction between samples. Statistical comparison of the content of samples obtained by each participant, regardless of surface, show some distinction between samples collected by Person 1 and Person 2, with some overlap, as seen in Figure 1B and 1D within the study. Overlap indicates similarities in microbial content between the samples obtained by each participant. When samples are viewed based on surface type, rather than by participant, distinct groups form for shoe and phone samples, with little overlap, as seen in Figure 1A and 1C within the study. This suggests that phone and shoe samples do not share many similarities in microbial content. Floor samples were tightly grouped with their respective shoe samples, indicating that the microbiome found on shoes is similar to the microbiome of the floor.

A random forest model was then generated to determine whether the participant that collected a phone or shoe sample could be predicted, allowing the authors to evaluate the investigative power of microbiome in this scenario. A random forest model generates a decision tree based on multiple variables to predict the classification of a sample, e.g. phone or shoe. The authors reported that random forest models were capable of correctly predicting which participant the sample originated from. Correct determinations for shoe samples were provided over 50 times as effectively as would be expected by random chance. An assessment of the microbes present in the shoe samples revealed 100 taxa, i.e. groups of organisms, which were determined to be significant. The authors report that the majority of the 100 taxa were consistently detected on one participant’s shoe samples, but not on the other participant’s. The differentiation of participants by shoe samples was attributed to the difference in the 100 taxa. Similarly, when phone samples were investigated by the random forest model, each participant could be differentiated. No distinguishable differences were found between the four shoes areas that were sampled, nor between the front and back of the phone, for each individual at each sampling time.

A random forest model was then applied to determine which microbial taxa were most associated with each surface type. Shoe and floor samples merged into one class, due to similarities in microbial composition. When trained at the genus level, shoe/floor and phone samples were distinguishable. Skin associated genera, such as Streptococcus and Corynebacterium, were more associated with phone samples compared to shoe/floor samples.

Samples were compared at different time points to determine the variability in a sample surface and sampling location throughout the day. Shoe samples were reported to be similar at all four sample locations at each time point, even as the microbiome changed across time points, indicating that all shoe sole surfaces possess similar microbial content at any given time. This suggests that changes in the microbiome of the shoe are rapid and spread quickly, since the microbiome of the shoe sole is relatively consistent across the surface at each time. Conversely, the front and back phone surfaces were not consistently similar at a single time point as the microbiome changed over time. While the microbial content of the samples was similar overall, the rate of change in microbial content between the front and back were not the same. The authors stated one likely cause being that the back of the phone possesses microbes sourced primarily from the hand, while the front has microbes sourced primarily from the face. Additionally, hand associated microbes have been reported to be highly volatile, i.e. prone to evaporation. The authors theorized that the volatility combined with the low abundance of the microbes on the phone surface may contribute to the differences in the phone samples at a selected time point.

The authors then expanded their research to include a larger pool of participants. Shoe and phone samples were collected from a total of 89 volunteers at three academic conferences, each in a different city. Subsequent data analysis showed significant differences between the microbiomes of phone and shoe samples. Both shoe and phone samples grouped significantly by location. Visual representations of the groupings can be seen in Figure 4 within the study. Random forest models were generated and were capable of predicting which city the sample originated from for both shoe and phone samples; correct predictions were significantly greater than would be expected by chance. Shoe samples were observed to provide more correct predictions than phone samples in these models. The authors stated that the results support the idea that different locations have significantly different floor microbiomes, enabling the samples to be localized.

Scientific Highlights

  • Shoe and phone samples were successfully distinguished from one another, through statistical models, based on the microbial composition of the samples.
  • Samples from participants 1 and 2 were distinguished from one another, through statistical models, based on the microbial composition of the samples.
  • Samples obtained at academic conferences held in three different cities were differentiable based on the microbial content of the samples, with samples separating based on location.
  • The rate of change for phone microbiomes was more variable than in shoe samples. Shoe samples showed consistent similarities in microbial content at each time point, indicative of rapid homogenization of the bacteria on surface.


Research has indicated a potential for microbial transfer, i.e. the transfer of microbes from a person or place’s microbiome, to be used to establish a link between an individual and place, object, or other individual. In this study, the rate to change in the microbiome on an object, i.e. phone or shoe, as well as the differentiability of objects and individuals based on microbial content were examined.

Potential Conclusions

  • Microbial content of a sample may be used to determine the geographical origin, i.e. location, of the sample or the individual that the sample originated from.
  • The ability to determine location based on shoe samples may be useful only if the sample is collected immediately at the site, due to the rapid changes observed in the microbial content of shoe samples, and therefore would have little forensic use.
  • Phone samples may provide individualizing information that would be of forensic use, due to the slower rate of change in microbial content. This may imply that other objects, such as keyboards or doorknobs, have the potential to provide individualizing evidence, if the microbial rate of change is similar to that of the phone samples.