Category: College of Engineering and Computing

1
00:00:00.830 –> 00:00:16.510
Mahin Chowdhury: Alright, Hello! My name is Mahin Chowdhury, and this and this is ‘The Purification of DNA Origami Nanoparticles’. Firstly, I would like to thank Dr. Remi Veneziano for helping me with with this project from start to finish, and I would also like to thank the Oscar program for providing me with this wonderful opportunity.

2
00:00:17.650 –> 00:00:37.060
Mahin Chowdhury: So the overview of the project: DNA origami nanoparticles is a new vaccination method that represents the next generation of vaccine delivery to fight infectious diseases. The current method towards purifying these nanoparticles relies upon ultracentrifugation, using Amicon Ultra filters which typically result in a low yield

3
00:00:37.060 –> 00:00:38.860
Mahin Chowdhury: In a laboratory environment this is-

4
00:00:38.940 –> 00:00:40.710
Mahin Chowdhury: -standard, pretty all right.

5
00:00:40.760 –> 00:00:57.070
Mahin Chowdhury: However, if you were to scale this up to the industrial level, it becomes a lot more complicated, as this will increase costs to manufacture and produce the nanoparticles. So the main objective is to find a way that that that could help reduce the cost or

6
00:00:57.100 –> 00:01:00.760
Mahin Chowdhury: purify the the structures as much as possible

7
00:01:02.340 –> 00:01:08.960
Mahin Chowdhury: The methods that we will be using: So we will be using 3 different DNA nanoparticles of

8
00:01:08.980 –> 00:01:22.690
Mahin Chowdhury: 3 different sizes, basically a 6-Helix Bundle, which is referred to as 6HB. Pentagonal Bipyramidal (PB), and Pentakis Dodecahedron (PD). As I mentioned, before, these particles are different than one another.

9
00:01:22.730 –> 00:01:33.020
Mahin Chowdhury: They have different properties, shaped differently. There are some similarities between 6HB and PB in terms of size (base pairs).

10
00:01:33.360 –> 00:01:36.170
Mahin Chowdhury: Pentakis Dodecahedron, on the other hand, is a lot bigger.

11
00:01:36.460 –> 00:01:43.420
Mahin Chowdhury: So we will be trying to purify these 3 types of nanoparticles, using the methods shown below

12
00:01:44.740 –> 00:01:59.890
Mahin Chowdhury: The first method we use is the regenerated cellulose membrane filter. As you can see, we have the experimental concentrations retrieved from the trials and the theoretical that was calculated, based on the volume of a retrieved at the end of the trial.

13
00:02:00.240 –> 00:02:10.600
Mahin Chowdhury: As you can see, there are some differences between the theoretical and the experimental concentrations. I encourage you that you please pause and look through these diagrams.

14
00:02:10.800 –> 00:02:18.500
Mahin Chowdhury: These diagrams / data would be compared to the agrose gel that we collected via a gel electrophoresis.

15
00:02:18.740 –> 00:02:32.360
Mahin Chowdhury: These bright bands are the nanoparticles. These bands below are the staple strands, or any byproducts that may come off from purifying the actual nanoparticles.

16
00:02:32.430 –> 00:02:36.960
Mahin Chowdhury: These are what we’re trying to get rid of essentially, as you can see.

17
00:02:37.010 –> 00:02:43.120
Mahin Chowdhury: The brightness indicates the concentration of both the nanoparticles and the staple strands.

18
00:02:43.570 –> 00:02:49.990
Mahin Chowdhury: So the brighter it is, the more concentrated it is. Meaning, there’s more of it; there’s a lot in the actual solution.

19
00:02:50.750 –> 00:03:06.030
Mahin Chowdhury: As you can see the 30kDa, based on the each structure. You can see that each band becomes fainter and fainter as you go up the scale for the filters, which indicates that the filters

20
00:03:06.220 –> 00:03:13.570
Mahin Chowdhury: do purify the staple strands or byproducts as much as possible.

21
00:03:13.870 –> 00:03:15.570
Mahin Chowdhury: Moving onto trial 2.

22
00:03:15.650 –> 00:03:21.770
Mahin Chowdhury: We can see that it is also a little bit consistent in comparison to trial one, and

23
00:03:21.900 –> 00:03:25.660
Mahin Chowdhury: if you compare it to the gel from electrophoresis

24
00:03:25.700 –> 00:03:35.150
Mahin Chowdhury: you can see that these bands become faint or barely visible, which shows that it has become pretty pure, and

25
00:03:35.150 –> 00:03:52.780
Mahin Chowdhury: could often indicate that some based on the comparison with the theoretical and the experimental, we can determine whether or not we lost the majority of the nanoparticles, or we retain them, in which case some we lose a lot, and some we aren’t able to purify as much

26
00:03:53.010 –> 00:03:57.120
Mahin Chowdhury: Moving on the Spin X polyethersulfone filters.

27
00:03:57.260 –> 00:04:05.080
Mahin Chowdhury: These filters are used in the similar method with the Cellulose filters. As you can see-

28
00:04:05.670 –> 00:04:25.000
Mahin Chowdhury: -you have the 6HB. Which is similar in the theoretical and experimental. However, if you were to compare it to the agarose gel, you have a lot of stable strands, so you can’t always rely upon the numerical data / figure. You have to use both the data and the visual gel.

29
00:04:25.190 –> 00:04:31.400
Mahin Chowdhury: In which case you can see that many for, trial one at least, many of these are not that pure

30
00:04:32.030 –> 00:04:39.690
Mahin Chowdhury: Moving onto trial 2, using the same gel, it’s pretty consistent, not as pure as it would seem.

31
00:04:39.740 –> 00:04:43.120
Mahin Chowdhury: Especially for PD, since the bands are very bright.

32
00:04:44.190 –> 00:05:01.580
Mahin Chowdhury: The Spin Kleen columns by Bio-Rad. They have a very high theoretical concentration, but a low experimental, which typically indicates that we we lost a lot more of the nanoparticles, and, as you can see here.

33
00:05:02.160 –> 00:05:06.100
Mahin Chowdhury: we actually have a a lot more nanoparticles

34
00:05:06.320 –> 00:05:15.770
Mahin Chowdhury: band of nanoparticles, but not a band for the staple strand, which means that we were able to purify it. However, at the cost we lost a lot of product.

35
00:05:16.660 –> 00:05:28.630
Mahin Chowdhury: The same is consistent with trial, 2, in which case trial 2 also had some consist of the same consistency as trial one. We lost a lot more of the nanoparticles, but we were able to purify as much as we could.

36
00:05:29.440 –> 00:05:49.230
Mahin Chowdhury: Moving on to this Zbba spin desalting columns, we could see that the experimental concentration is really high compared to the theoretical concentration, and this is also backed up by the bands. The staple strand bands that we see here, especially for PD. Which are pretty bright, indicating that these are not as pure as it should be.

37
00:05:50.520 –> 00:05:54.140
Mahin Chowdhury: The same thing goes with the trial 2 same consistency.

38
00:05:54.500 –> 00:06:01.490
Mahin Chowdhury: It just shows that these are these columns are not a viable method towards purifying the nanoparticles.

39
00:06:02.640 –> 00:06:04.330
Mahin Chowdhury: So the conclusions

40
00:06:04.380 –> 00:06:22.040
Mahin Chowdhury: Based on the data, many different nanoparticles are best purified by a specific method. PD, for instance, is very well purified, using the 100kDa cellulose membrane filter

41
00:06:22.050 –> 00:06:26.750
Mahin Chowdhury: There is still some access that you may have, but

42
00:06:26.880 –> 00:06:44.460
Mahin Chowdhury: you still have a good amount of nanoparticles. The same goes with the other nanoparticles. They all have a different type of purification yield, depending on whether or not the method, the method, whether filter or column, can purify, based on its structure and size.

43
00:06:44.780 –> 00:06:56.690
Mahin Chowdhury: so as they as dated for some methods would just result in a setting pure structure, but with a high yield. Other structures

44
00:06:56.770 –> 00:06:59.780
Mahin Chowdhury: have a very pure structure, but a low you.

45
00:06:59.810 –> 00:07:02.570
Mahin Chowdhury: So is this optimal for industrialization?

46
00:07:03.150 –> 00:07:04.560
Mahin Chowdhury: Not as much.

47
00:07:04.630 –> 00:07:17.860
Mahin Chowdhury: It could be better, but it is mainly because we just don’t have a very viable method at this point right now. The current method is still the best method.

48
00:07:17.920 –> 00:07:24.710
Mahin Chowdhury: On the other hand, we were able to find that many different structures are purified better with different methods.

49
00:07:24.740 –> 00:07:30.860
Mahin Chowdhury: better way to purify each structure based on a different method that we can use.

50
00:07:31.050 –> 00:07:44.890
Mahin Chowdhury: So what’s next? The next? The next step is to continue searching for a better better solution. This is not the end. This is merely the beginning of finding a better way to purify these nanoparticles.

51
00:07:45.020 –> 00:07:50.950
Mahin Chowdhury: depending on the size or shape. we just have to find something that works for that specific nanoparticle.

52
00:07:53.250 –> 00:08:06.870
Mahin Chowdhury: Thank you so much for listening to my presentation. Please leave any questions if you have any. Once again. Thank you so much to Dr. Veneziano and the Oscar program, and thank you again for listening to my presentation. I hope you have a great one.

So starting off by talking about how telescopes are calibrated, what you need
is a reference star. Basically this is a base of all your measurements, and what you do is you can measure the light coming out from the Star and then you
can compare it to what is known for the reference star. So if they match, then your telescope is calibrated properly, and if they don’t match you can adjust
your settings.

So currently all SI traceable astronomical measurements are based on
measurements from one star named Vega that were taken in the 1970s which is a pretty long time ago. With modern astrophysics you need a
lot more precision and accuracy, so NASA and NIST need a new method to calibrate their telescopes so that way they can have more accurate measurements. Some other methods that you can use to calibrate a telescope is using an artificial star which is basically a very strong laser that’s also very
precise, and this way you can use it to act as a reference star instead of using
an actual star so this way you can control the light intensity. The goal of my project is creating a very high precision laser controller that will be flown on the spacecraft and will act as an artificial star to calibrate telescopes, and my goal is to have a very low percent error less than 0.25 percent. So here’s a block diagram of my full system so we have the laser driver which is supplying current to the laser itself and then we have the TEC controller
which is used as thermoelectric cooling which will control the temperature and keep it steady for the laser. Then we have two ADCs and then we have it all connected to a Raspberry Pi.
Taking a closer look at the laser driver, it has five volt DC power and
then it has these inputs so you can control the levels of the light intensity of the laser and current level. Then it has all these outputs that we’re going to track.
It’s the same thing with the TEC controller, so we have the temperature setting, the voltage setting, and then we have all of these outputs that will
tell us it’s working properly and that will sense the temperature and the
current.
Here’s a photo of my first prototype that I’ve been testing. This one has
the laser mock-up on it, so the two diodes right here and then it has a fan,
a heat sink, and then it has the TEC module which will maintain a steady
temperature right underneath the heat sink, and you can kind of see it in the
side view but the two wires here the red one and the black one are for the TEC. Maintaining steady temperature is really important to keeping the error
low so I designed a quick experiment to make sure that the TEC was working
properly. So the way the TEC works is that it can either heat or cool depending on the direction of the current, so if your current is forward then it’s heating and if your current is reversed then it’s cooling. I had three trials with three different current levels 0.25, 0.5, and 0.75 A, and then I tested how well the TEC worked for all of them.
Basically I measured how long it took for the temperature to peak and then I
measured the maximum and the minimum temperatures for each one as you can see it here in the plots. For the first trial with 0.25 amps we have the temperature increasing up until about 30 degrees Celsius where it peaked,
and then I turned off the supply at 100 seconds which is when the peak time was, and then you can see it cool back down to room temperature. Then for reverse current, the orange one, it does the same thing but this time it’s cooling and then heating back up. For trial two it looks about the same, but the temperature time was a lot faster at 60 seconds this time versus
100 from before and the peaks are also a lot taller. Same thing for trial three with 0.75 amps, so this time peak temperature is at 58 seconds and the maximum and the minimum temperatures are a lot higher
than they were before.
Basically that tells us that TEC works properly and you can control the amount of current to control how fast you want it to heat up or what your
temperature is desired to be. Right now what I’m still working on is putting everything together and interfacing it with Raspberry Pi. I’m working on doing breakout boards like this one that you can see here, that will make
sure each chip is working properly before I put them all together.
Finally some applications of my project are calibrating NASA and NIST telescopes as well as the GMU telescope. They can also use this for dark
matter calculations and finding new exoplanets and black holes.
Thank you so much for watching!

Hello! My name is Zan Stuart, I’m a Senior Mechanical Engineering Student and today I’ll be talking about my project, which is a mechanical and surface characterization study of 3-D printed PLA-HA Composite.

Slide 2:

So the motivation for this project is the materials currently used for hip and knee joint replacements, such as stainless steel and titanium, are adequate, but they have some significant issues, such as the fact that they’re expensive, they’re difficult to machine into these complex shapes that are necessary for this application, and they can cause some adverse long term side effects such as stress shielding, which basically wears away the bone material around the replacement joint.

Polylactic acid, or PLA, and Hydroxyapatite, also known as HA, composite is a potentially viable alternative. PLA is a very common plastic used for 3-D printing, and hydroxyapatite is a mineral that’s commonly found in your bone structure.

PLA-HA is pretty cheap, it’s biocompatible, and it’s easily adapted for 3-D printing.

Very few studies exploring the properties of PLA-HA in general exist, and fewer, if any, study the properties of 3-D printed PLA-HA.

Slide 3:

To briefly state the objectives, we wanted to manufacture the PLA-HA composite using different manufacturing methods, we wanted to process that material into a form that useable with a 3-D printer, and we wanted to 3-D print some samples and test their properties, and study how the manufacturing method effects the properties of 3-D printed PLA-HA.

Slide 4:

Two different manufacturing methods were used, one is dry speed mixing, which basically involves placing the raw PLA pellets in a container, mixing them with hydroxyapatite powder, and spinning it very very fast in a dry film until the mixture is roughly homogenous and then turning that into filament.

The other is magnetic stirring, where you basically dissolve the PLA pellets in a solvent, and then put them in a magnetic stirrer, which mixes them up as you can imagine, mixing that viscous dissolved PLA mixture with hydroxyapatite powder and then a silane coupling agent and then forming that into raw PLA-HA after it cures, which you can see in the middle of the slide.

Slide 5:

That raw material is then placed into a filament extruder which you can see on the right side of the slide here, where you put the raw material in the hopper up top, it is then compressed and fed through a heated nozzle at the end where it spits out something resembling a plastic wire which is then fed to a 3D printer.

Slide 6:

Here is just a basic schematic of how 3-D printing actually works, like I said, the plastic string once again gets fed through a heated nozzle, the bed upon which the part that you’re building is formed on moves around along with the nozzle so you get the proper geometry as you desire.

Slide 7:

Here are a couple of examples of the completed samples I made for this project. On the right is a magnetic stirred sample as well as a failed one, and here’s an example of a dry speed mixed sample.

Slide 8:

Two different testing methods were used, one was microindentation which I won’t cover in extreme depth here, but basically it involves pressing a very small diamond tipped pin into the surface of the material, and you use the data collected from that to observe the mechanical properties. In this case we wanted to find the elastic modulus and contact creep performance.

Creep performance is a measure of how much a material deforms with time with a constant load. As you can imagine that’s something you want to keep to a minimum. And the SEM image here is just an example of one of the indentations that was performed on a magnetic stirred sample.

Slide 9:

The other method used was Energy Dispersive Spectroscopy or EDS. Here’s the basic working principle of it, you’re basically shooting an electron beam at the surface of the sample, that excites the atoms and displaces electrons at lower energy states. Following the principle of conservation of energy, when an outer shell electron moves to fill that vacancy, because it’s at a higher energy state that excess energy has to go somewhere, in this case it goes to X-rays. These X-Rays can then be received. Each element actually emits a unique X-Ray during this process, so by figuring out what the wavelengths of the x-rays are and receiving them we can figure out what the composition of the sample is.

Slide 10:

Here are the results of EDS that you can actually see, the highlighted areas or the colored areas indicate the presence of the element given on the left. Hydroxyapatite is composed of calcium oxygen, phosphorus, carbon, and hydrogen. Calcium is not present in the base material of PLA, so if we see calcium or phosphorus we know we’re seeing HA.

We can see that HA was actually integrated successfully using both manufacturing methods. As far as the concentration goes it’s hard to say much, as the area that we imaged was very small. However, this had to be done because the hydroxyapatite powder used is only visible on the nanoscale, so it’s hard to make a solid observation other than the fact that we know that HA is present in the PLA material.

Slide 11:

And here are the results of the mechanical testing, you can see that the contact creep was significantly higher for the magnetic stirred samples which is less desirable, and the elastic modulus of the dry speed mixed samples was significantly higher than that of the magnetic stirred samples.

A two sample t test was conducted to determine the significance of manufacturing method on these two properties, and you can see that yes, the manufacturing method significantly influenced both properties.

Slide 12:

In conclusion, the manufacturing method significantly altered both the elastic modulus and creep performance, and hydroxyapatite was successfully integrated into the PLA matrix using both manufacturing methods.

If I was to continue this project in the future, I would produce and test more samples, and use different manufacturing methods such as wet speed mixing and re-extrusion.

I would also conduct more and different mechanical tests, such as wet and dry wear tests to see what the friction properties of the material are, as well as tensile tests to more thoroughly evaluate the elastic modulus.

Slide 13:

Thank you for your time and attention, and I hope that you have a great day!

With this in mind, I wanted to find a way to determine a minimum amount of input data needed to make an effective whitening filter.

The plan was to create and test whitening filters using the following method:
• First, generate correlated data.
• Process it through an ideal whitening filter.
• Decompose the correlation matrix of filtered data into eigenvalues and eigenvectors (more on those later).
• Plot the eigenvalues, and then observe the results.

In more detail:
Step 1 is simulating correlated data.
• I started with MATLAB’s random normal generator—the uncorrelated data from this function represents white noise.
• Then I created two correlating functions—a sinc function for noise coming from three dimensions and a Bessel function for noise coming from two dimensions.
• I built correlation matrices for each.
• I broke those matrices down into Eigenvalues and Eigenvectors, which in this case can be thought of as convenient tools for doing matrix operations.
• Then I used these tools to transform the data.

And at that point in the process, the data represented correlated ambient ocean noise.

As an example of correlation vs. non-correlation, here on the left is a scatterplot of two vectors (one on the horizontal axis; one on the vertical axis) that are correlated. You can see that as the values for one of them increase, the values for the other one increase as well, and that’s why you get kind of a slanted plot. On the other hand, on the bottom right you have a more circular pattern, and that is a scatter plot of two vectors of uncorrelated data. And then finally on the top right is a plot of one column of the correlation matrix of sinc-correlated data, which is actually what I used in my project.

After correlating the data, I tested it using the inverses of the transformation matrices, which are actually the ideal whitening filters. I passed the 3D data through a 3D whitening filter and the 2D data through a 2D whitening filter.
Here are graphs of correlation matrix eigenvalues of a single trial for various sample sizes for each data type. So, to understand these graphs, you need to know that eigenvalues represent the weights—or amounts—of different spatial frequencies present. You can see that the graphs flatten out as sample size increases. And this means the spatial frequencies are becoming more evenly distributed, which is the goal of a whitening filter.

Finally, to evaluate my findings, I wanted to take a closer look at the eigenvalues of the correlation functions of the filtered data.
Here I have plotted 1000-trial histograms of 3D data passed through a 3D filter showing the distributions of Eigenvalues with varying numbers of data snapshots. You can see that the Eigenvalues for small numbers of snapshots are spread out. This means the data is not effectively whitened. But as more snapshots are added, the eigenvalues cluster more closely around one. That is the desired outcome for a whitening filter.

In conclusion:
I can see there’s a threshold where the eigenvalues begin to converge around a single value, but I am not yet certain where it is. There’s more to study, and I plan to continue this project.
Hopefully continued research on this topic will lead to improved whitening filter performance.

Here are references I consulted during this research and my acknowledgments. Thank you very much for watching.

In this video, I will first explain what PEEK/HA is and how it can be applied. Next, I will describe the filament extrusion process and different manufacturing methods of the material. Lastly, I will provide my results, difficulties, and progress since last semester.

*
PEEK/HA is a composite material composed of polyether ether ketone and hydroxyapatite. In short, a composite material is a combination of different materials with the intent of combining or improving specific attributes.

PEEK is a high performance polymer with characteristics such as, but not limited to: biocompatibility, non-toxicity, radiolucency, and a bone-like elastic modulus. One of its few drawbacks is its inherent biological inertness. Supplementing it with an additive such as HA, however, reduces PEEK’s inertness and can encourage bone growth.

This in particular opens up the possibilities of utilizing PEEK/HA in orthopedics, and, given its plasticity, even 3D printing bone scaffolds and implants. This could be used to repair bones as you can see in the image on the left. A basic, but real 3D printed PEEK scaffold is shown on the right.

To use PEEK/HA material in 3D prints, the composite material is often made into pellets and drawn into filament using an extruder such as this one. Pellets are dropped into the funnel, where an auger pushes them further into the barrel, where they are melted down and pushed out of the nozzle as a thin filament.

The first manufacturing method I employed to create PEEK/HA material was dry mixing. This method involves mixing the PEEK pellets with HA powder one to two times at 3000 rpm for 30 seconds. I would then extrude the pellets into filament.

The second method, re-extrusion, is really just an extension of dry mixing. Basically, I would create the dry mixed filament but then cut it into pellet sized pieces and run them through the extruder one to two more times to better mix the composite material.

In situ was the final method I employed. Unlike dry mixing, the in situ method chemically forms PEEK/HA in one go. As such, it is the most complex and time consuming of the three. It involves adding various compounds into a heated mixture and gradually increasing the temperature until the solution is formed. It is then poured out into a tray and broken into extrudable pieces.

Each method aimed to improve the thoroughness of the PEEK/HA mixture at the cost of time and simplicity. The original goal of my research was to 3D print mechanically testable samples using each method, but given the inherent difficulty of extruding and printing with a high melting point thermoplastic, I spent most of this semester just trying to print dry mixed samples.

The image on the left details cylindrical PEEK compression samples which I printed for practice before using the composite material. This allowed me to get a better understanding of the intricacies of 3D printing with PEEK before beginning the process with PEEK/HA. The image on the right details my two closest attempts at creating rectangular prism PEEK/HA compression samples. The one on the left failed during printing due to a section of the extruded filament diameter being too large for the 3D printer to use. The one on the right experienced a similar problem in addition to gaps in the layers likely due to the problematic diameter again or uneven cooling.

Overall, practice with regular PEEK filament and utilization of better adhesive techniques brought me one step closer to printing a complete and testable sample compared to the previous semester. However, when I continue this project this coming summer, I might consider using an automated filament winder to obtain a more consistent diameter. I also want to investigate the effects of speed on the quality of the prints as well.

Thank you.

What are DBPs?
Disinfectant byproducts or DBPs for short are chemical compounds that form when disinfectants in water react with natural organic matter.
Chronic DBP exposure can cause significant negative health effects, such as bladder cancer, colon cancer, and pregnancy complications.
A person can be exposed to DBPs through highly disinfected water sources, for example, chlorine interacting with organic matter in water. To ensure drinking water safety, it is imperative that DBP levels in public water sources be properly monitored.

Unfortunately, DBP exposure is difficult to measure directly. Instead, epidemiologists have been using Trihalomethanes or THMs, which are easier to measure, as a surrogate for DBP exposure. This was because it was believed that THM concentrations are proportional to concentrations of other DBP classes.
A previous study examined the link between THMs and a DBP class Haloacetonitrile or HANs using over 9500 measurements from 248 public water systems. This study found that THMs could only explain 30% of the variance in HAN concentrations.

For the project, we wanted to create a statistical framework to model the concentrations of hard-to-measure unregulated DBPs that drive toxicity using a wide array of co-occurring DBPs.
We also wanted to take into account the public water system (PWS) of origin for the measurements.

The data used is from the Information Collection Request database from the environmental protection agency or EPA. The dataset has more than 13,000 measurements from 295 public water systems.

For models, Linear mixed models or LMMs were used. LMMs are useful for data with high variability between groups. And in this specific case, the variability between public water systems can be considered.
LMMs allow estimation of the fixed effects (ex: DBP concentration or categorical variables) that can be measured while accounting for the variability among groups (ex: Public Water System) with random effects.

For variable selection, LASSO regression was used, which allows us to select important variables using a penalization approach with a tuning parameter, lambda.
as you can see on the top graph here as we increase lambda the coefficients tend towards 0
and in the bottom graph you can see how we select lambda, by running LASSO regression with multiple lambdas and measuring the mean squared error, which is a metric used to evaluate prediction accuracy.
We select all DBPs that have nonzero coefficients.

Multiple model structures were tested against the selected model structure (Full LMM)
The metrics used to evaluate models included AIC and BIC which are for comparing Goodness of fit (lower is better)
Conditional R-squared which measures the proportion of variance of the response explained by the model. (higher is better)
RMSE which is for comparing prediction accuracy.

We found that for four different DBP classes (haloacetic acids-5 haloacetic acids-6, haloacetonitriles, and haloketones), we can model a significant amount of the variance with a unique group of DBPs when we take into account the water system and key categorical variables.
In the upcoming summer, we plan to conduct an in-depth missing analysis of data and prepare new models for better estimation of the concentrations using the information gained from this model. We also further plan to develop a methodology to classify at-risk water treatment systems.

This is my work cited and thank you for listening to my presentation.