Loading...

Software

Diagnostic Software

See our team’s GitHub Judging Release for our finalized code of our software: Click Here

To help with diagnosis, we wanted to provide a yes/no vote from each biomarker as to whether the patient has sepsis. For our three cytokines (IL-1beta, IL-6, and TNF-alpha), the diagnostic cutoffs reported in the literature are values in plasma, but for CRP and Lactoferrin the values are reported in serum. Luckily, for CRP there is no reported difference between the levels in serum and plasma (for two different types of anticoagulant used)¹, so we can compare it directly to the cutoff. For Lactoferrin, however, we multiply the plasma concentration resulting from the previous two equations by 1.41 (the ratio of the means in control patients)² to obtain the serum concentration and compare this to the cutoff value for a yes/no sepsis vote. Because our experimental data on resistance vs. sweat concentration measures those concentrations in nanomolar, nanomolar is the unit in which any equation obtained from our Calibration Curve would output sweat concentration. As such, we converted the diagnostic cutoffs found in the literature for each biomarker into nanomolar units using molecular weights found on the Genscript website. Our software compares the predicted blood concentration (serum or plasma, depending on biomarker) to these converted cutoffs.

Biomarker	Optimal Cutoff Value (variable units)	Converted Cutoffs (nM)
IL-1-β	0.33 pg/mL³	0.000157
IL-6	16pg/mL³	0.00076
TNF-ɑ	5 pg/mL³	0.00029
CRP	80 mg/L⁴	.67
Lactoferrin	675 ng/mL⁵	8.44

Table 1: Diagnostic cutoffs reported in the literature for sepsis for each of our 5 biomarkers. For all our biomarkers, it is above the cutoff that indicates increased risk of sepsis. The conversions to nanomolar were done using molecular weight from the GenScript website.⁶

Our software first runs the measured resistance values through our electrical to sweat concentration conversion model, and then takes the data from this conversion, and runs it through our second model, which converts sweat to blood concentration. For each biomarker, if the level predicted by the equations determined from our modeling is above the diagnostic cutoff, the software displays “Sepsis risk.” Physicians will of course look at quantitative diagnostics, like heart rate, as well as run other tests to inform their decision-making, but our data can be crucial for them to diagnose sepsis more accurately and faster.

Assumptions

The data in Barthe et al. on serum vs. plasma levels of Lactoferrin was taken on control patients. We are implicitly assuming that the ratio of Lactoferrin from plasma to serum would be the same in sepsis patients as it is in the control patients.²
As explained above, we assume that CRP levels are the same in the two media.

Testing

Figure 1: Values of plasma concentration based on random inputs of resistance to simulate the function of our software, as well as the yes/no Sepsis Risk reading output by our software. Note that concentration and time values are not representative of what one should expect, they are only for demonstrative purposes.

Ease of Use

We commented on the software in a way that it can be used by any other team easily for biomarker data that they have. We made provisions to include different equations to convert biomarker concentration from blood to any other media provided modelling equations are known and we provided graphing ability within the code. We also made it possible to change the disease being tested for and provide thresholds for each biomarker. In order to actually test the ease of use, we asked the Osiris RIO UFRJ iGEM team to use their voltage values for biomarkers with our software and gained some very valuable input to make the software more user friendly. Their feedback helped us improve the way we commented on our code thereby improving ease of use.

Future Directions

We can’t yet do real time monitoring via the potentiostat, but the data obtained from the potentiostat can be used with the software. We use the potentiostat data (voltage and current) as input and then output a graph of the predicted plasma concentrations of each biomarker in the simulated “patient,” and give a corresponding readout of yes/no Sepsis Risk for each biomarker. (Figure 1)

Here we are indicating sepsis risk if any of the biomarkers are above the known thresholds. This is a straightforward reading intended to give physicians extra information, but we would be excited to work on a more nuanced diagnostic model incorporating the concentrations of all biomarkers relative to their cutoffs into one probabilistic readout. We investigated doing this by creating a combined Receiver Operating Characteristic (ROC) curve. This would tell us the combined True Positive Rate (TPR) and False Positive Rate (FPR) for our five-biomarker-panel, but the necessary data was not available in the literature. Specifically, to combine ROC curves, we would need to know the diagnostic cutoff that corresponds to each (TPR, FPR) data point on the ROC curve for each biomarker, but the authors to whom we reached out for this data about their published ROC curves did not reply. Other exciting future directions for incorporating all biomarkers into one reading involve Decisions Trees and Bayesian equations. The latter would require further research into the extent to which elevations in our biomarkers are independent from each other as opposed to upregulated by each other.

References

Aziz, N., et al. Analytical Performance of a Highly Sensitive C-Reactive Protein-Based Immunoassay and the Effects of Laboratory Variables on Levels of Protein in Blood. Clinical and Diagnostic Laboratory Immunology. 2003, 83 (9), 652-657. DOI: 10.1128/cdli.10.4.652-657.2003
Barthe, C.; Galabert, C.; Guy-Crotte, O.; Figarella, C. Plasma and serum lactoferrin levels in cystic fibrosis. Relationship with the presence of cystic fibrosis protein. Clinica Chimica Acta. 1989, 181, 183-188. DOI: 10.1016/0009-8981(89)90186-1
Angeletti, S.; Dicuonzo, G.; Fioravanti, M.; De Cesaris, M.; Fogolari, M.; Lo Presti, A.; Ciccozzi, M.; De Florio, L. Procalcitonin, MR-Proadrenomedullin, and Cytokines Measurement in Sepsis Diagnosis: Advantages from Test Combination Disease Markers. 2014. DOI: 10.1155/2015/951532
Sierra, R., Rello, J., Bailén, M.A.; Benítez, E.; Gordillo, A.; León, C.; Pedraza, S. C-reactive protein used as an early indicator of infection in patients with systemic inflammatory response syndrome. Intensive Care Med. 2004, 30, 2038–2045. DOI: 10.1007/s00134-004-2434-y
Huang, K.; Du, G.; Wei, C.; Gu, S.; Tang, J. Elevated serum lactoferrin and neopterin are associated with postoperative infectious complications in patients with acute traumatic spinal cord injury. Arch Med Sci. 2013, 5, 865–871. DOI: 10.5114/aoms.2013.38680
Reliable Services & Products to make Research Easy. GenScript.

Team:Rochester/Software