A new way to diagnose lupus

Harnessing the power of Nanosensors and Machine Learning to package the various Lupus biomarker tests into one, increasing the speed and accuracy of Systemic Lupus Erythematosus (SLE) diagnostics.

(Written by Ashley Mo, Marzooq A, Isabelle Lau)

80 years ago, the first case of Cystic Fibrosis (CF) was recorded, and little did we know that this small abnormal gene would cause thousands of deaths to date. The good news is, we can currently get ahead of CF by detecting it early on and treating its symptoms.

Today, the 70,000+ individuals living with CF are expected to live a normal health span. And this is all thanks to medical advancements in diagnostics.

Unfortunately, this is not the case with Lupus.

Lupus is an autoimmune disease in which the immune system is unable to distinguish harmful, foreign invaders from healthy cells. In response to healthy cells, the body produces antinuclear antibodies (ANAs) which form nearby nuclear antigens and lodge themselves in organs or blood vessels causing inflammation, or latch themselves onto the receptors of innocent cells to be destroyed.

Retrieved from CreakyJoints.org

This causes a large range of complications to the body, the most notable being kidney failure, blood disorders, butterfly rash, and inflammation of the lungs. But there is currently no cure for lupus, and so lupus patients can only be treated for their symptoms, not lupus itself.

Lupus has earned the infamous nickname of the disease of 1,000 faces because of its ability to show a range of illnesses all over the body while manifesting itself differently to everyone. Some people may exhibit only 5 recognized lupus symptoms whereas others are photosensitive or have rheumatoid arthritis. This also means that common symptoms of lupus can also be common symptoms of other diseases.

Lupus diagnosis rates are only reaching 50%.

Note that this percentage could be higher because there are likely to be more people unknowingly living with Lupus. This means doctors are trying to solve the wrong problem by giving them the wrong treatment.

Although Lupus can be drawn back to a combination of environmental and genetic factors, its root cause is still unknown. It is highly probable that there isn’t a common root cause to be found, since lupus manifests itself uniquely to each person.

There are over 1,000 known lupus gene susceptibilities that have various levels of strength. The gene susceptibilities alone may or may not be harmful, nor do they guarantee an onset of lupus. They are only activated when DNA becomes so damaged (typically through sun exposure or certain drugs/medication) that cells undergo apoptosis (programmed cell death) and nearby immune cells recognize the cell’s destroyed DNA as external foreigners and begin producing ANAs.

This is why lupus diagnosis is so difficult.

Since there is no common root cause, currently diagnosis methods of lupus rely on a combination of very unspecific symptoms. According to the Lupus Research Alliance, at least 4 out of 11 common lupus symptoms must be noticed, so strongly manifested in the patient that it disrupts their day-to-day activities.

Not only does it take time for symptoms to be visible, doctors may or may not order all possible lupus biomarker tests. This includes ANAs, blood disorders (lack of red & white blood cells and platelets), other proteins produced as a result of inflammation, etc. This often leads potential lupus patients down rabbit holes of a seemingly never-ending series of tests upon tests.

This causes multiple problems:

  • Time-consuming — the tests are only taken one by one and are often conducted over a 12-hour time period. It can take up to a week for patients to hear back.
  • Costly — one antibody/antigen/protein test can cost up to $60 USD
  • Must be done by a doctor — tests often require access to labs, patients need to book multiple appointments with rheumatologists which can take up to several months.
  • Are subject to human error — since most test results are observed by a naked eye

This is partly why it takes nearly six years for a person to be diagnosed with lupus, on average. The consequences resulting from tedious, inaccurate and constant methods to track lupus biomarkers only elongate patients suffering by delaying them from receiving proper treatment to suppress their painful symptoms.

Introducing Helio: The world’s fastest and most accurate lupus diagnostic test

Helio uses next-generation aptamer-based biosensors, big data sets, and machine learning to determine if a patient has lupus — all within a few minutes.

Biosensor

The biosensor is made from aptamers as our receptor component and a graphene transducer.

By using an aptamer, we can detect and differentiate between types of ANAs and red blood cells due to their high specificity and affinity. When the biomolecule binds onto the aptamer, oxidation and reduction reactions cause a movement of electrons which is detected by the graphene transducer.

Each aptamer is engineered to the specific biomolecule which will ensure high selectivity for us to differentiate different types of ANAs and red blood cells.

Big Data Sets

We can combine information collected by the nanosensor into a data set that is unique to the patient. Hidden patterns as well as information about the onset of lupus is better interpreted through a large data set. This also allows tests to show more quantitative data (numbers, percentages, concentrations etc.) versus qualitative data (descriptions, perspectives, observations) and give a more accurate diagnosis of lupus.

Machine Learning

Helio uses a supervised artificial neural network to pursue the diagnostic process. The model is primarily a statistical analysis method used to predict a data value based on prior observations of a data set. Based on historical data about earlier outcomes involving the same input criteria, it then scores new cases on their probability of falling into a particular outcome category.

It is one of the most common ML algorithms that is already being applied to diabetes prediction and cancer detection.

This is roughly what type of output we expect from the aptamer biosensor. The various coloured lines will represent the various proteins/antibodies/antigens discovered through binding and their electrochemical responses will be shown as a graph.

This information is taken and fed through the algorithm as an input, the output will produce a final percentage stating the likelihood that the patient has lupus (since a definitive response from the algorithm doesn’t consider the outward symptoms of lupus)

The model is trained with data that we collect from people who are confirmed to have lupus, those who don’t have lupus and some who may exhibit lupus symptoms but don’t have lupus. Through supervised learning, the model will learn to recognize the results of patients who likely have lupus and we will use an additional random assortment of datasets to further validate it’s accuracy after training.

A faster proper diagnosis isn’t far away.

Without a common root cause for each patient, combining all the biomarker tests into one and using a machine-learning algorithm to make a diagnosis is timely, inexpensive and accurate.

Our goal is to turn those months of testing into just a few minutes, bringing the correct treatment faster to those who truly need it.

Taking the misdiagnosis rate of Lupus from 50% to 0% is ambitious, but at Helio, we believe that it is possible.

A misdiagnosis shouldn’t be what stops someone from living their best life.

As the world gets closer to a cure for lupus, we are doing our part, starting with a proper diagnosis of lupus. Our solution is currently under validation, but we know this technology is something our team is determined to bring to reality.

Because healthcare is about living a healthy life, regardless of race, gender, sexuality, religion, disorder or disease. Not about late treatments, an endless series of tests, and apologetic doctors.

If you would like to hear more about Helio’s research, ask questions or leave comments, feel free to contact us. We would love to hear from you!

Ashley: www.linkedin.com/in/ashley-mo-a044781b5

Isabelle: https://www.linkedin.com/in/isabellelau123/

Marzooq: https://www.linkedin.com/in/marzooqa/

16 y/o learner