BME100 f2018:Group17 T1030 L6

From OpenWetWare
Jump to navigationJump to search
Owwnotebook icon.png BME 100 Fall 2018 Home
People
Lab Write-Up 1 | Lab Write-Up 2 | Lab Write-Up 3
Lab Write-Up 4 | Lab Write-Up 5 | Lab Write-Up 6
Course Logistics For Instructors
Photos
Wiki Editing Help
BME494 Asu logo.png

TEAM 17

Name: Enedino Sosa
Role(s)
Name: Haleigh Hunt
Role(s)
Name: Jacob Hershkowitz
Role(s)
Name: Sierra Wilferd
Role(s)


Our Brand Name

LAB 6 WRITE-UP

Bayesian Statistics

Overview of the Original Diagnosis System

BME100 tested patients for a Parkinson's disease-associated SNP. 17 teams of 4 to 5 students diagnosed 34 patients. The work was divided by assigning 2 of the 34 patients to each of the 17 groups. In order to provide consistency and limit errors throughout the groups, three replications of each patient's sample were ran through the PCR machine along with positive and negative control groups, similar imaging stations were set up for analysis, and the analysis software used was the same for each group. Calibration tools for the imaging stations all used calf thymus DNA concentrations.

Each of the patients' samples were divided into three smaller samples, which were used to create a solution for the PCR machine. This enabled us to ensure that our results were consistent with the other, and provided us with a "back-up" in case something went wrong with one of the solutions. The negative and positive controls allowed us to observe any discrepancies in the process, as anything but a negative (no-DNA) and positive (yes-DNA) respective result would have alerted us if something was incorrect. The imaging station was the same throughout each group, and the standards were the same. The imaging software, ImageJ, was used by each group to determine the positive or negative results for each of their PCR samples. Each sample used three droplets for consistency purposes, and the means of these droplets were used for diagnostic purposes.

Discrepancies may arise due to the variability in phone cameras or distances of the phones from the droplets. Droplet size may also vary between groups. The PCR samples may have also been contaminated with other DNA, and, in some cases, the amount of samples used to create the PCR samples were not enough.

The results for each patient are described below.

Patient ID PCR Conclusion Doctor Diagnosis
64002 NO TEST yes
85406 NO TEST no
15385 POSITIVE no
95785 POSITIVE yes
49652 NEGATIVE no
99085 NEGATIVE yes
41967 POSITIVE no
46547 NEGATIVE no
26433 POSITIVE no
39343 NEGATIVE no
23643 POSITIVE yes
47311 NEGATIVE yes
28063 POSITIVE yes
67655 NEGATIVE no
12210 POSITIVE no
21105 POSITIVE no
17649 POSITIVE yes
28137 POSITIVE no
28319 POSITIVE yes
81362 NEGATIVE no
55425 NEGATIVE yes
75247 POSITIVE no
58152 NEGATIVE no
69578 NEGATIVE no
38209 POSITIVE yes
96033 NEGATIVE no
79929 NEGATIVE no
84911 NEGATIVE no
40686 NEGATIVE no
68520 NEGATIVE no
76880 POSITIVE no
89675 INCONCLUSIVE yes
38060 NEGATIVE yes
54209 NEGATIVE no


What Bayes Statistics Imply about This Diagnostic Approach


Calculation 1 is the chance that a patient's test result will be positive, and that it is an accurate conclusion(the patient actually does have the SNP, and the PCR test results say so). i.e., there is a "this chance" that the individual was properly diagnosed with a true positive.

Calculation 2 is the chance that a patient's test result will be negative, and that it is an accurate conclusion (the patient actually does NOT have the SNP, and the PCR test results say so).i.e., there is a "this chance" that the individual was properly diagnosed with a true negative.


Calculations 3 and 4 are the chance that a patient's test result will indicate the development of the disease. They describe the probability of the patient developing the disease indicated by a positive test result and the probability of the patient NOT developing the disease indicated by a negative test result, respectively.


Three possible sources of human or machine/device error that could have occurred throughout this experiment include contamination or improper pipetting techniques when preparing the PCR samples, not using the correct sample amounts for the PCR preparation and/or for creating the droplet, and not setting the fluorescence device up the same way for each droplet. The last is especially true when using different phones and using different distances of the phone from the droplet while imaging. Contamination of the PCR samples could result in false negatives or positives, improper pipetting would result in inconsistent data and possible PCR failure, and the inconsistencies in imaging would result in a misinterpretation of data. Each of these factors would skew our Bayes values in different directions.

Intro to Computer-Aided Design

3D Modeling

Group 17 used SolidWorks in order to design our new fluorimeter machine using computer-aided design. SolidWorks allows us to view a 3-dimensional rendering of our product, and to manipulate different dimensions to achieve the accuracy in imaging we needed to create a more stable product. Each individual piece was designed around capturing the image of the drop consistently and accurately. We no longer wanted the camera or fluorimeter to move, yet still wanted to show how the ceiling could open up for easy access to sample placement. SolidWorks allowed us to create using the exact dimensions for the finished product, and provided us with a drawing that can be used during production.

Our Design


  • Images here not to scale

G17 Four.png Image 1: Closed box; "In-use"; cords not shown
G17 Three.png Image 2: Inside of the stabilized box with camera block and fluorimeter light
G17 Two.png Image 3: Camera
G17 One.png Image 4: Hinged door of the box



This design provides the user with ease of access to the glass slide on the fluorimeter light, as shown in image 2. Image 4 shows how the door would attach and move between light blocking and access-open positions. Images 2 and 3 detail how the camera would fit into its block, stabilizing it and providing consistent images. The entire unit is built together in order to prevent the camera focusing and position changes we experienced during our procedure (image 1).


Feature 1: Consumables

Our product is a redesign of the fluorimeter machine we used in order to make it more friendly to class lab applications. It provides uniformity in measuring the DNA values of each droplet, thus providing consistency throughout a group experiment. Due to this, our machine requires all of the consumables used in lab 4 that were used to run the PCR machine and place droplets on the glass slide (PCR reaction mixes, tubes, pipettes, ect) to be provided by the user. ImageJ is still used so that each student can manipulate their data. A "how-to" guide covers the importance of proper pipetting techniques and how to use the fluorimeter with ImageJ.

Our machine does require connection cords, which are built in to the box and provided.

Feature 2: Hardware - PCR Machine & Fluorimeter

The OpenPCR machine is still used "as-is" in our design. We exchanged the fluorimeter for our "Fluore-Sense" fluorimeter. Our fluorimeter provides more consistency in imaging, leading to more overall consistency in sample comparison and diagnosis. It also has easy set-up procedures, simply needing to be attached to a power supply and a laptop for image transference. Our imaging system does require some technical "know-how," as the images are meant to be immediately transferred to a laptop. A file for the image transfer, as well as monitoring to maintain sample ID matching with the images is required.


The major weaknesses in the fluorimeter were the inconsistent distances from the droplet to the camera, different cameras (phones) being used, and the different specification of the camera being used. Our machine limits this by stabilizing everything into place. The only thing that moves is the ceiling to the black box so that a sample may be easily transferred onto the glass slide. The glass slide is securely fit into a slot, and the camera is stabilized on an immovable side of the box. The camera and camera specs do not change based on group, and the distance is always the same from camera to droplet. Images are transferred directly to a laptop for processing, limiting data loss. User error is reduced, and more consistent results are obtained.