Circuit Modeling

Overview

      According to our genetic circuit, the output of our test kit will be green fluorescent or red fluorescent signal corresponding to GFP or RFP creates. However, in some sense, color is subjective. Fluorescent light might not be easily observed by human eyes, not to say those who have color blindness. In this case, the users might read the results incorrectly and the test kit will be wasted. Therefore, we find it necessary to develop an app complementary to our hardware package so that the results can be more readable and interpretable. In addition to that, to suit diverse background of users, a detailed menual on our Cereulide Testing Kit and educational information related to synthetic biology will be provided in our app.

Methods

      In our app development process, there are basically three steps. First, we trained a machine learning model which is going to be used in the Color Identification function. Our model is trained using Google Teachable machine[1]. Next, we designed the prototype of our app. Finally, we'll program our app on Android Studio[2] according to the prototype design we made so it will be able to help the users to analyze the results of the test kits with their mobile devices in the real world.

Color Identification Machine Learning Model

      As we aim to identify whether the output fluorescent signal is in green or red color, we plan to train a machine learning model using Google Teachable Machine to do color identification. Google Teachable Machine is a platform where users can create their own machine learning model easily by providing a dataset containing inputs and labels of inputs. To train our model, we are collaborating with hardware to collect photos of gels with fluorescent protein inside and divide them into the Green group and the Red group accordingly.

      Once the training is done, the model will be able to output the probabilities of test case belonging to each class. The class with the highest probability will be the final results on the surface of our app.

      Google Teachable Machine provides the functions for user to enter the input and do the real-time classification on their webpage. However, we want to integrate the whole classification process in our app. Therefore, we exported our model generated from Google Teachable Machine and programmed it into our app in later steps.

App Prototype Programming

We have done the programming of the prototype on Android Studio, which is the official Integrated Development Environment (IDE) for Android app development. Here is the current interface we have developed:

Figure 1 Home page of our app
Figure 2 User can choose to upload their photo for result analysis
Figure 3 User could also take the photo from the app directly

      After that, the user could click the 'Predict' button to generate the result. As we do not have a training dataset now, we haven't built a model yet (check Color identification machine learning model for future plan) . Below is the demo of how the result would looks like:

Figure 4 Analyzed result as unsafe
Figure 5 Analyzed result as safe

Future Directions

  1. Sign in to save the test results as a record.
    1. Include data recording of event result with date and time for future tracking of the result
  2. Calibration of the phone camera.
    1. As every phone camera is different, there would be differences in the quality or colour of the image, which might lead to incorrect results when passing the input from the users to our machine learning model.
    2. Before use of the app, users would need to do a calibration test by taking the object we would provide with the test kit. We are going to do calibration of the camera of every user before they use the app. We would take a photo with our camera (the camera that takes the sample photo of the gel to be put into the machine learning model) and compare it with the user calibration photo from their camera. Then adjust the variation of every input photo that is due to their camera quality before passing them to our trained model.

References


    [1] Teachable Machine, https://teachablemachine.withgoogle.com/ (accessed Oct. 12, 2023).
    [2] Android Studio, https://developer.android.com/studio/intro (accessed Oct. 12, 2023).