• Capital Expenditures
  • About
  • Contact
  • Capital Expenditures
  • About
  • Contact
CAPITAL EXPENDITURES

a learning investment in DAta Science, entrepreneurship, and Biotech

by
​ ​Vanessa Mahoney

MATHEMATICAL MATTERS OF THE HEART

9/30/2016

1 Comment

 
Picture
Figure from my paper showing electrical activity in scar cells. Panel A: my experimental setup for recording electrical activity. Basically I loaded the hearts with voltage sensitive dye so I could record movies and SEE electrical activity. The black circles correspond to the pixel plot traces that are shown in B, and the drawn circle represents the scarred region. Panel B: Pixel plot tracings show communication between scar cells and the surrounding cells. These "hills" we see in the pixel plots indicate that the cells are depolarizing. Translation: there is electrical activity across that scar, which is occurring because the cells can communicate with one another. In C and D, we removed a protein we hypothesized was responsible for providing this interaction, and you can see many of the lines are flat, indicating no current is flowing through these cells.
For my post today, I want to direct you to the blog of my brilliant former colleague, Gary Mirams, a computer science research fellow at Oxford (click here to read his post on this subject.) Gary helped support my PhD research by convincing non-believers that the findings I gathered in mouse hearts were real, and in fact supported by math. (Or maths, as they say across the pond. )

My research investigated electrical interactions between cardiomyocytes (the primary working cell in the heart) and fibroblasts (the cells that come in to repair the heart after injuries such as heart attack). More specifically, I recorded electrical activity across the injured regions of mouse hearts, as you can see in the figure above. What did I find when I looked at the scars? Electrical signal! Using several different methods -  again and again -  we saw that there was electrical activity in those scars. This was pretty exciting, but also controversial: the canonical view is that cardiomyocytes are the only cells that are capable of relaying electrical current. But if we see electrical activity in the scar, and we know there are not cardiomyocytes in the scar, then it follows that fibroblasts can conduct electrical signal, too.

​This is an important finding, because it would change the way we treat patients. Fibroblasts populate scars after heart attack, crowd the spaces between myocytes as we age, and proliferate like crazy in stiff, overworked hearts (think heart failure). If these cells are conducting signal, instead of being impediments to signal, this could drastically alter the way we treat the arrhythmias seen in each of these cases.  If fibroblasts are actively contributing to arrhythmia development - being part of the wayward circuit - then it is possible that if we shut down that activity, we could shut down those errant arrhythmias. 

But before it comes to treatment, it is necessary to verify the results of investigative science. We had a reason to look for signal in mouse hearts: there a multitude of scientific studies and clinical observations that indicate it is possible for fibroblasts to conduct electrical activity. However, none of the studies or observations were waterproof: one could always point to an explanation that weakened the  "fibroblasts can communicate, too" contention. My study attempted to show, incontrovertibly, that in the intact heart, fibroblasts can conduct.

We showed evidence of electrical activity and the absence of myocytes in the heart several different ways, through several different experiments and repeats of experiments. However,  we only seemed to arouse more questions in the disbelievers. And you can't blame them; it is common medical practice to go create a scar in the heart to interrupt arrhythmias. (But patients need to come back up to 50% of the time.... which argues our point.) 

Anyway, enter dazzling pundit, Dr. Gary Mirams. People didn't believe that the signal we saw in intact hearts was real, so we said let's go see what happens if we model this electrical behavior with computational simulation using what we knew about the electrical properties of the cast of cardiac cells. 
What we knew:

-  the dimensions of the mouse heart
- the dimensions of the scar
- the electrical properties of myocytes
- most of the electrical properties of fibroblasts and other cells in the scar
What we didn't know:
- how densely packed the cells in the scar were
​- how well those cells communicate with one another

Those two things we didn't know have a large effect on the coupling (electrical activity that is possible) in the scar. So, being scientists, we gathered data, made assumptions, tested a range of feasible assumptions, and ran models for each of them. The result: Gary could replicate the findings I found in mouse hearts without having to make any crazy assumptions.

This is where I will direct you to Gary's post. He shows in detail the specific equations that were used to model a wave of electrical activity moving across a scarred region. If you scroll down you'll see the videos that mimic what we saw in my mouse hearts, with Gary's comprehensive explanation of what you are seeing, why you are seeing it, and what changes if we adjust those unknown properties. I admit I'm a nerd, but it is so cool to see how you can break down a complex biological process, learn from the equation before you even get started, and make a video simulation of a cardiac electrical wave. (Note, find the actual videos of  electrical activity across mouse hearts in the supplemental information section of my paper.)

I also stole Gary's blog title for this post. So perfect. Thanks, Gary. :) 



Sources:
1. Gary's blog: Mathematical Matters of the Heart
2. My PhD paper
3. Al Green: How to Mend a Broken Heart 

.
1 Comment

SORE TODAY, STRONG TOMORROW 

7/31/2016

1 Comment

 
Picture
​It was Amazon Prime Day this month, and being an immutable spendthrift, I couldn’t help but rummage for a bargain. My one purchase was an aggressively large tub of whey protein (discounted 50%)! Before I pulled the trigger, I conducted some brief Google Scholar research on the effectiveness of protein powder. Is there compelling evidence that protein powder helps non-elite athletes build muscle better than other forms of protein? No, not really. But I bought it anyway. There isn’t evidence the stuff hurts you (if you don't overdo it) and as a protein, it is filling. Moreover, working out feels good, and if you can extend that feeling by putting something in your body that is intended to improve your body, than why not do it? And if protein supplements encourage you to keep up your resistance training, then aren’t they - in effect - helping you build muscle? So rather than criticize the claims of women’s magazines or expound upon the studies showing the inadequacy of protein supplements as muscle builders (because I kind of want to), my stance is go for it. 

But still being interested in directly building muscle, what can I do, especially as a woman? Levels of testosterone, which is a steroid hormone produced naturally by Leydig cells, are significantly lower in women than men: we have just 20-30% the amount, a significant biological disadvantage when it comes to building muscle mass. Muscle building happens after muscles are broken down, or hypertrophied. Testosterone enhances this hypertrophy, and because of the increased muscle breakdown, there is consequently more effective rebuilding of the muscle. (Anabolic-androgenic steroids replicate this natural muscle-building effect of testosterone.) Increasing testosterone levels naturally is not very effective in spurring muscle growth, and I’m certainly not advocating steroid use, either. So what’s a girl to do?


Like many things in life, there’s no quick fix. The key to muscle building is commitment and effort. It takes time: one study cited that the very earliest muscle mass increase can be seen is three weeks following a dedicated resistance training regimen. But even before the cross-sectional area of the muscle increases, the muscle architecture actually acclimates. In other words, your muscles start physically gearing up before the actual amplification starts.

That’s not the only adaptation occurring with resistance training: neurological connections between your brain and the muscles are also being enhanced. In fact, following the commencement of resistance training, electromyography (EMG) recordings show increased muscle activation, suggesting that the initial gains in strength are due to neural factors. So before your muscles get bigger, your brain is more effectively recruiting motor units to be involved in the motion. In fact, one study showed that thinking about muscle contraction – without actually doing the physical work – can modulate physical strength! Two healthy groups of individuals underwent immobilization of the wrist, with one of the groups being instructed to think about muscle contraction for the week of immobilization, and one taking no such step. The group that thought about moving their wrist showed significant increases in wrist strength! It was only a week, and "not losing as much strength" isn’t quite the same as "gaining strength". However, mental imagery is known to stimulate cortical areas that are involved with motor behaviors - so who knows -  perhaps the effects could be more significant and prove more far-reaching over time.


Well, since I’ve spent this post being a buzz kill, I’ll end with some positive muscle-building evidence.  Multiple studies have shown that overloading the muscles (read, repetitions until failure) combined with stretch is the most effective natural stimulus for promoting muscle growth. Stretch+overload builds muscle by adding sarcomeres (the basic unit of muscle), both in parallel and in series.  See the chart below for the types of activities that cause the greatest relative recruitment of motor units. Looks like I’m only using about 20% of my muscles rowing, but CrossFit moves like Clean and Jerks and Snatches recruit upwards of 75%. That doesn't mean you should stop doing isolated lifts - in fact, it is easier to build muscle from targeted, upper body muscle resistance training, likely because there is more direct neural activity involved. So get lifting! 

Picture
Sources (Coming Soon)
1.
2. 
3. 
4. 
5. 
6.
​7.
1 Comment

Ice, Ice Baby

6/27/2016

0 Comments

 
Yesterday as I was icing my plantar fasciitis-stricken heel for what feels like the millionth time, I began wondering what sorts of sports recovery technologies on the horizon might help get me past this despised duty. A recovery technology that has been getting a lot of celebrity and athlete press lately is cryotherapy, or ultra low temperature therapy. Not for just freezing off warts, local and whole body cryotherapy is now proposed to be useful for muscle pain, muscle metabolism, performance, depression, anxiety, and acute and chronic inflammation, among other things.   
Picture
Ice bath vs cryotherapy! Image from OhioCryo
Picture
​Like many “health” trends today, there is little evidence that cryotherapy is effective. I’m not saying it’s ineffective –  my skeptical scientist perspective just recommends that you should not rush to go shell out $90 to stand in a cryo chamber for 3 minutes  - which is the rate at Kryolife in NYC. Does that sound ridiculous? Well there’s more: they recommend 10-20 treatments.  The misleading stacked bar graph in their brochure (below) implies cryotherapy patrons should expect reductions in pain, but further examination shows that the graph is only showing a self reported pain scale from 9 patients with Bechterew disease (a chronic inflammatory disease). A more explicit graph would show:
  • Number of patients (it’s not clear only 9 patients were tested)
  • What type of patients were tested, ideally from a broader subset  (because Bechterew disease is a pretty specific population)
  • Comparison against a control group – because the placebo effect is often quite significant.
  • A different data visualization – stacking bars on top of one another is purposefully deluding
This example illustrates my problem with cryotherapy: it could be effective, but where is the data? When data is presented like this, and it feels like you’re being swindled -  it makes you wonder why there wasn’t good data.

​I found a pretty interesting scientific review of whole body cryotherapy (WBC) that examined evidence from 10 different controlled trials to determine the efficacy and effectiveness of WBC. While some of the studies were limited and not entirely conclusive, only one of the studies showed WBC helpful in rehab – adhesive capsulitis of the shoulder rehab.  


To me, the most interesting figure in this journal article is the performance indicators graph. While reported pain is subjective and very much mental, measurement of quantitative performance is absolute. This forest plot summarizes different performance metrics of people who underwent WBC and people who did not. The outcome effect ratio is demarcated with the green line, while the vertical black line shows the interval or range of outcomes, with 95% confidence. Forest plots may look complicated, but they're actually pretty easy: think of the horizontal line as the representation of the test group. If the horizontal line touches the vertical line, the results aren’t significant. Why? Because this group falls within both the left AND the right side of the outcome effect ( with 95% probability). That means a subject in the group could be on the left - benefits from WBC - OR on the right: benefits from nothing. We can't reach a conclusion. However, if the horizontal line is completely in a camp, we have a 95% probability of knowing one of the treatments indeed changes the outcome. We see from this figure that one of the performance indicators actually was completely on the left side: Tennis shot accuracy is improved with whole body cryotherapy! 

Picture
Forest Plot of Performance Outcomes
So it’s not a lost cause, WBC  could be beneficial, but remember these studies just compare WBC to control treatment, which is active recovery but no cold therapy.  There are several other ways to cool the body, including ice packs and water baths. While cryotherapy is certainly the coldest method– at about - 250 degrees Fahrenheit it’s the coldest temperature in the world – that doesn’t mean it transfers cold to the body the most effectively. The poor thermal conductivity of air means that despite establishing the largest temperature gradient, the ability to lower the subcutaneous and core temperature of the body is relatively weak. This makes sense – being in 40 degree water will get your teeth chattering a lot quicker than 40 degree air.  So although WBC benefits from holistic cooling, it looks like crushed ice (which benefits from phase transfer) and ice baths can get the body just as cold - or even colder. 
​
So, I guess I’ll begrudgingly go grab a frozen water bottle for my heel; no cryotherapy for me.
 
Sources:
​1. ​Kryolife
2. ​ Studies on Whole-body Cryotherapy
3. Open Access J Sports Med: Whole-body cryotherapy: empirical evidence and theoretical perspectives
4. The Ice Treatment Cometh: Super-Cold Cryotherapy Arrives in CT
5. NY Times: Why Ice May Be Bad for Sore Muscles
0 Comments

The spreading of the flu... and false information

5/25/2016

0 Comments

 
Picture
Picture
It's been a dreary spring in New York, but with Memorial Day approaching, we finally (hopefully) have some warmer weather to look forward to! I hope you emerged from winter , perhaps a little whiter, perhaps a little plumper, but overall happy and healthy.  It is winter sickness, in the form of the flu, that got me thinking (something I do once in a while). 

I don't know how many times this winter I heard a little fallacy floating around:  that the flu shot causes the flu.  In fact I heard it in so many forms:  "I got the flu after getting the flu shot", "if you didn't get the flu after the shot then you're protected", "my friend got the flu after the shot but it's much less potent than the full blown flu"  that I started to wonder if this year's flu, like HBO's take on the Game of Thrones books, was changing the rules.

But no: no the flu shot can NOT cause the flu. A simple Google search shows, in a variety of reputable forms, that no, no you can't - or are highly, highly unlikely - to get the flu from the shot. 

Flu vaccinations are most commonly prepared by using an 'inactivated'  form of the virus. Translation: the infectious part of those little Vs is dead. Recombinant technology and a cell-based nasal spray are alternate preparations, but each of these technologies has only one FDA approved vaccine on the US market.   In randomized, blinded studies, where some people get flu shots and others get salt-water shots (placebo), the only differences in symptoms were increased soreness in the arm and redness at the injection site among people who got the flu shot. There were no differences in terms of body aches, fever, cough, runny nose or sore throat. What does that mean? It means people from both groups probably thought they had the flu, but there was no increased amount of "flu" in the group that got the shot. 

So what's going on? What's the disconnect? Why do almost half of Americans think  that the shot can cause the flu?  I think it comes down to these three things:


1)Humans are naturally inclinatined to find patterns. The flu shot is only ~ 65% effective. So sometimes if you get the shot, you still could get sick.  People have a natural tendency to see two events and assign one as the cause, and the other as the effect. This suspected causal link may look even stronger if someone gets the flu almost immediately after the shot. This does happen, but the shot wasn't to blame. The flut shot isn't actually effective for about two weeks - it takes time for antibodies to develop in your body. The incubation period for the flu is 1-4 days. What that means is if you're exposed to the flu during that two week period, you could absolutely contract the flu, because you weren't protected yet. Coincidental: yes. Causal: no

2) The internet can be a vehicle for the spreading of disinformation. When I want to go learn about something, my PhD diploma requires me to go look at scientific journals and government approved sites before I form my opinion (I'm being facetious, but you get my point). But because ANYbody can post almost ANYthing, there is a ton of crap out there.  I think an interesting side effect of the connectivity of the internet is that instead of informing and spreading truth, the internet can actually entrenching people in their ideas. If a falsehood is published, and it resonates because it happens to align with people's uninformed ideas of things, is said by a person of importance, sounds really good, would explain something that you heard once, etc, then it can actually make clarity of the truth MORE difficult to achieve, and make people MORE convinced that their wrong ideas are in fact correct.

3) Americans are deeply distrustful of medicine and science. It's a confusing world out there: big pharma is price gouging (that sounds wrong), GMOs are bad for you (we don't want to eat science, right? what does GMO means?), shots are giving our children autism (heard about that, must be true), people are overdosing on pain killers (doctors must be to blame)....  while I'm oversimplifying to prove a point, I can understand why a level of distrust has arisen.  We should monitor what goes in to our bodies, and there are some causes for public concern, even if they are placed there by the media or other's with something to gain. But many people seem to believe the whole medical and scientific community is out to get them.  I think what this comes down to is  people want to be in control. Perhaps it's not that people want to believe medicine is a conspiracy theory or a ploy to get money into someone else's pocket, but just that they have a desire to understand. And in trying to do so, perhaps in complex situations it's easiest to point the finger at something one doesn't fully comprehend.

I think all three of these phenomenon influence the way we form opinions, not just about the flu but the millions of things we are exposed to every day. There's a ton of noise, but we're each just trying to find a signal in it. :) 



Sources:
1. CDC: How influenza (flu) vaccinations are made
​2. Flu vaccine effectiveness: questions and answers for health professionals 
3. Reuters: Nearly half of Americans think flu shot can make you sick
4. GOT Gif.. just because 
0 Comments

Theranos: The Lance Armstrong of blood diagnostics?

2/22/2016

0 Comments

 
​As the opening metaphor suggests, Theranos, once the darling of biotech, has recently come under considerable scrutiny, similar to the prodigal cyclist caught doping to gain an edge. (Credit to Forbes for developing and expanding upon this analogy.)  While it’s not yet absolute that Theranos is guilty of perjuring the methodology and results of their platform to rapidly test a few drops of blood, this $9 Billion valued company is at the least guilty of misleading the public on the maturity of this technology. When this WSJ article surfaced accusations last October, Theranos CEO Elizabeth Holmes went from the cover of Forbes to an exec under fire; holmes' trademark reticence no longer seemed so eccentrically unique. While Holmes' immediately refuted the WSJ claims as false, look what did happen.....the Theranos website  noticeably changed. 
Picture
Theranos overstated it's ability to test with it's a "nano" amount of blood.
​The idea supporting the conception of Theranos is noble: blood testing is too expensive, too slow, and too painful, so here’s a technology to rectify all of that. As you can see below, canonical blood testing is fairly cumbersome: use a needle to puncture the arm, take at least a vial of blood, aliquot the samples to individual wells, load the samples to a mass spectrometer, and analyze and report the results. Theranos pioneered a technology called Edison that would perform a slew of diagnostic blood tests with just a finger prick, offering drastically reduced costs and processing times. The key to this platform is microfluidics, the miniaturization and automatization of regular blood testing. 
Picture
However, a series of journal articles, failed partnerships, employee testimonials, and even a suicide have cast considerable doubt on the veracity of Theranos’ purported revolutionary testing method.  The most salient arguments cornering Theranos are summarized here:
  • the Edison technology is not actually being used to do most blood testing
    • by the end of 2014 less than 10% of tests were using this technology
    • there are claims that tradition Siemens AG machines were used to report results
  • Many customers have reported grievously erroneous test results from the Edison testing
    • Potassium levels that would only be possible if “you were dead”
    • Safeway executive received Theranos results that suggested prostate cancer. Retesting by another lab showed normal results to the antigen.
    • One patient, two different results. Theranos results pictured are largely out of normal range. 
  • Theranos testing results have not been properly evaluated 
    • Theranos didn't have to prove to the FDA that their tests worked because they don't sell Edison technology to other labs.   
    • There are allegations from former employees that Theranos' president and COO ordered lab personnel to stop using Edison machines for proficiency testing and report only results from  instruments bought from other companies.
    • Theranos' self reported clinical correlations are missing statistics and number of samples, the omission of which is unacceptable in reporting scientific results.
  • A very small blood sample is problematic for diagnostics
    • Because of the small volume, contamination of the sample occurs much easier
    • Some tests require larger volumes, so the sample would need to be diluted, another step that alters the sample and is likely to introduce error
  • Pricking the finger breaks cells
    • Which can introduce fluids from tissues and cells
  • The demise of British biochemist, Ian Gibbons, who is the co-inventor of 19 of Theranos patents, suggests trouble at Theranos
    • Dr. Gibbons reportedly told his wife “nothing was working” at the company
    • Dr. Gibbons committed suicide in May 2013
  • Dissolved partnerships with Theranos portend more underlying problems
    • Walgreens demands answers to technology questions before it opens up any more clinics
    • $350 million Safeway-Theranos deal fizzled after Theranos missed deadlines for the blood testing roll-out  
While it certainly doesn’t look good for Theranos, there is still a chance that this growing company is simply unequipped to quickly mobilize in response to all of these requests. Is the technology too new? Probably. And as the pedal is pressed in innovation, tech missteps and prematurities will arise.  But what Theranos did do wrong is hide its errors. There is a problem when you can't share your results. Academia could not function without peer-review, accurate reporting, and eventually, collaboration. While I understand the need for tech companies to protect their intelligence, we must not forget to protect the system from corruption. 

That said, I think rapid testing of blood/fluids is a promising horizon.. While my limited research did not uncover widespread commercialization of microfluidic diagnostic technologies, exciting research out of The Methodist Hospital Research Institute, Rutgers University, and Wake Forest Baptist Medical Centre indicate commercialization could be forthcoming. Microfluidics and lab-on-a-chip technology is honestly pretty cool.  Look below to see how this V-chip could provide test results from up to 50 tests using ink and gas (a gross simplification. ) Again, these results came out of a lab, which is really a prototyping environment that does not guaranteed widespread success, but the principles that this V-Chip reside on are worth investing in. 


Picture


Sources:

1. Forbes: America's Richest Entrepreneur's under 40
2. WSJ: Hot Startup Theranos has Struggled with its Blood-Test Technology
3. Forbes: Is Theranos too Good to be True? 
4. Quantified Health: Theranos Unmasked 
5. WSJ: Safeway, Theranos Split after $350 Million Deal Fizzles
6. WSJ: Walgreens Scrutinizes Theranos Testing
7. Nature Communications: Multiplexed Volumetric Bar-Chart Chip for Point-of-Care Diagnostics. 
​
8. Science News: 'Lab-on-a-Chip' technology to cut costs of sophisticated tests for diseases and disorders. 
9. Theranos.com 
10: Tech Insider: Here's what we know about how Theranos' 'revolutionary' technology works 








0 Comments

Machine Learning 101

1/11/2016

0 Comments

 
Picture
​

There has been quite a lot of talk of machine learning and artificial intelligence in the media, a trend that is all-but guaranteed to continue. Machine learning implementations are everywhere: from consumer-facing examples like google searches, Siri speech recognition, and spam filtering,  to sophisticated commercial technologies like unmanned drones and self-driving cars. machine learning has infiltrated our daily lives, although not quite in the ways portrayed by Hollywood. 
​
People have disparate feelings regarding the imminence of artificial intelligence (AI). An observer article reported that technology cognoscenti Stephen Hawking, Bill Gates, and Elon Musk are deeply concerned with the risks of artificial intelligence, with Musk citing AI as “our greatest existential threat”. However, machine learning is also revolutionizing the diagnosis and treatment of disease. Check out this paper from the University of Sao Paulo, in which a machine learning algorithm is used to diagnose Alzheimer's Disease from electroencephalography (EEG) patterns, with both high accuracy and sensitivity! Personally, I think Spiderman said it best when it comes to the development of AI: "With great power comes great responsibility".  Specifically, there has got to be some regulatory oversight, both domestically and internationally, when it comes to programming machines to act intelligently. But rather than launch into an ethical discussion, today I'd simply like to unshroud some of the basic concepts of machine learning.  
A few data science terms: 

machine learning: a subfield of computer science that involves "teaching" computers to recognize patterns in data and apply these findings to make decisions.
supervised learning: the computer "learns" from input data and builds an algorithm that can be used to make predictions on a similar data set. 
unsupervised learning: the computer does not have labels for input data, and thus must structure the input data on its own in order to map outputs. 
cognitive computing: a computer system that seeks to mimic the human thought process. Cognitive computing uses machine learning algorithms, natural language processing, and data mining to continually learn, adapt to new problems, and model solutions. 
artificial intelligence: a broad computer science field which describes the intelligence exhibited by computers. Encompasses learning, representation, reasoning, and abstract thinking. 

Some common machine learning task by outputs:
classification: computer classifies new data (ie: is this incoming email spam? yes or no) usually by learning from input data 
clustering: computer divides data into groups without knowing what the groups are.
regression: statistical technique to determine relationships between a dependent variable and one more independent variables. 
Picture
Figure 1: Data Science Skill Sets and Subfields
To start, let's zoom out and see where machine learning is in the spectrum of data science. The above Venn Diagram breaks down data science into some easy to digest concepts. First, computer science can be viewed as broad skill set or field under data science, along with math & statistics and subject matter expertise. At the interface of these skill sets is subfields. For example, my graduate research in electrophysiology can be defined as traditional research. I utilized my subject matter expertise in the heart to design and perform experiments, and by applying statistical techniques to my findings, I was able to show that I had made significant findings. Now machine learning lies at the interface of computer science and math & statistics: machine learning relies on a computer program's ability to apply statistics to problems. (Side note: a "unicorn" data science is the holy grail data scientist, having expertise across the all three areas and rumored to make on average over $200k.) 

Machine learning certainly sounds like an intimidating term, but let's start with a very simple example of machine learning: k-means clustering. In this example, let's pretend a business wants to group online customers into similarly-behaving cohorts in order to deliver targeted marketing at scale. While a multitude of variables on the customer would be gathered (demographics, online behavior, affiliate connections) to keep things simple let's just look at two inputs: how much the customer spends on average, and how environmentally conscious the customer is. 
The following infographic, adapted from a post by Ben McRedmond, walks through the basic steps of k-means clustering. 

Picture
Figure 2: K-means clustering
In the last step, we can see that further iterations will not change the cluster assignments: we have found the clusters! Now the business can market to all of the customers in cluster 1 a certain way and all of the customers 2 in a different, more targeted way.

Some notes about k-means clustering: while k-means clustering will always terminate, the first random placement of the green points will affect the clustering. What that means is we could end up with a less than optimal set of cluster assignments based on that initial step. In order to minimize this error, k-means is often repeated to find the best cluster definitions.  Also, as I pointed out above, the cluster center is the average of the components, so outliers can skew the cluster by influencing the placement of the cluster center.

​This is just one example of clustering, but this type of analysis is used in a variety of other fields and applications. In biology, cluster analysis can be used to building group of genes with similar expression patterns, to make spatial and temporal comparisons of communities, and to differentiate between different types of blood and tissue in a three-dimensional image. This is just one simple example of a machine learning technique, and of course machine learning algorithms get much  more complicated, but I hope this illustrated the basic premise of machine-learning: giving computers the ability to solve problems without explicitly programming them.  

Sources:
1. ​The Observer: Stephen Hawking, Elon Musk, and Bill Gates Warn About Artificial Intelligence
2. Med City News: 
4 ways healthcare is putting artificial intelligence, machine learning to use
3. Forbes: The Hunt For Unicorn Data Scientists Lifts Salaries For All Data Analytics Professionals
4. Wikipedia: Machine Learning
5. Blog Intercom: Machine Learning is Way Easier than it Looks 
​
6. Wikipedia: Cluster Analysis
7.Clinical EEG and Neuroscience Improving Alzheimer's Diagnosis with Machine Learning Techniques
0 Comments

The patient as consumer: because if healthcare is going to stay expensive (it is), it better be quality

11/23/2015

0 Comments

 
Picture
Picture
In NYC, if you an ominous orange “C” appears in the window of a restaurant, you’re likely to redirect your patronage to an establishment that has scored higher on the sanitary grade scale. Dr. Raz Winiarsky, cofounder and chief medical officer of Spreemo, thinks like restaurants, medical provider quality should also be more visible and influential. The focus of Spreemo is radiology in workers’ compensations claims. (This isn’t boring, keep on reading!) Why this specific niche? The answer is twofold: 1) radiology is a medical service that is treated like a commodity: people seek the cheapest or most convenient option. However, diagnostic imaging from disparate providers is not like varieties of apples;  there is an expansive range in the quality of radiology, and as such there is a significant degree of patient outcomes based on the radiologist. 2) Spreemo has found a niche where it is actually feasible to get providers to compete with each other based on quality instead of price: workers’ compensation. Workers’ compensation is insurance that provides injured-on-the-job-employees with wage replacement and medical benefits, provided the employee relinquishes his or her right to sue the employer (Wiki). As such, these insurance companies desire to get employees healthy and back to work as soon as possible. And as it turns out, if you get employees better treatment, they are likely to return to work quicker. Spreemo’s Yelp-like star rating system for radiology quality is based on several factors including education and training, the quality of equipment, and price. Patients are rewarded with better quality, while radiologists adhering to quality guidelines can expect higher reimbursements.

This idea that healthcare should be quality driven is not novel, but in this country, it is far from operational. The fee-for-service model, which predominates today, reimburses healthcare providers based on the volume of services they provide, not the quality of care. And while I do not wish to disparage healthcare providers– because I do believe physicians as a whole place the well-being of their patients first – why  invest in more expensive equipment, training, and services if there is no incentive to do so? As Dr. Winiarsky contends, our system chases the cheapest options for care, effectively driving healthcare quality lower and lower. A radiologist on Spreemo's site voices: why buy a new, $1.5 million dollar imaging magnetic resonance imaging (MRI) machine when an old $200,000 piece of equipment will suffice, and there is no forward incentive to cover the $1.3 million differential?

Picture

​But the problem is, that $1.5 million piece of equipment likely enables the radiologist to see the pathology better. Furthermore, an MRI on its own is not a conclusion: there is an extensive spread between MRI diagnoses. In New York City alone, the disparity of care was shocking. Dr. Winiarsky’s team took patients to a dozen different radiologists to get MRIs (don’t worry, no radiation here) and observed not just a range of diagnoses, but wrong diagnoses. And a radiologist’s diagnosis isn’t an isolated measurement: it affects patient treatment, prognosis, and time back to work.

Healthcare is driven by money, but in this little niche, Spreemo has indirectly found a way for quality to be the driving factor. But can this work in other areas of medicine?  In other industries, the consumer has a powerful voice, as the customer’s reviews, patronage, and freedom to choose between competitors drives businesses to satisfy the customer. However in healthcare, the consumer not only expects poor service at a doctor’s office, they still almost always return to the same office. Furthermore, even though consumers are allowed to select their own health insurance, the choices are often all just different tiers from the same provider. And what incentive does the insurance company have to deliver good service? The patient pays for services up front (because you select a plan and are locked in) so there is no incentive to foster patient-insurer relationship, loyalty, or quality of interchanges. 

But the patient as consumer is a revolution that is slowly approaching. Patients are interacting with the healthcare system much more frequently. Startups like ZocDoc allow patients to review doctors, while Medisafe is an app that allows users to track medication dosage adherence. In addition, social media outlets and wearable fitness devices mean patients are generating and viewing  much more data about their health. Moreover, the recent shift to high deductible health plans incentivizes consumers to shop around and make informed decisions for the best plans, For example, the WSJ reported that during this season’s healthcare enrollment season, many employers introduced computer tools to help recommend the best plan for each employee. 

the consumer not only expects poor service at a doctor’s office, they still almost always return to the same office.
Picture

There is another push for healthcare consumerism, and it comes from costs. Healthcare costs in the United States are high, and while there are proposals that could bring down the cost of healthcare (eliminating fee-for-service), costs will in all likelihood remain high. These costs could be kept in check by the government, but socialized medicine (socialism in general) is not something Americans easily stomach (Bill Maher and Bernie Sanders recently debated socialism on Real Time). Many Americans are uncomfortable with the idea that the government can give and take something away, especially something like healthcare. So if then healthcare is not treated like an entitlement, but rather a commodity, then the price of healthcare will be dictated by what the market can bear. In other words, the cost of a medication – say a medication that is the only approved drug for a disease – will be set at the highest point at which it will still be purchased. One can see the enormous hazards of such a handling of healthcare. However, if healthcare is still treated as more of a commodity, but it is possible for the patient-consumer to have a voice and a say, then perhaps medical care prices can be commensurate with the consumer’s regard for the medical care provider.  High costs could at least be justified with quality behind it. A consumer should be able to choose expensive, best-in-class services, or have the freedom to select a less expensive option when basic treatments will suffice. In this system, healthcare is still being treated like a commodity, but in a way that empowers the patient consumer. 

​Healthcare consumerism is dawning, but it has not yet arrived. Startups like Spreemo are pioneering change, making us realize old principles of healthcare are flawed and need to be recast. In order for healthcare consumerism to truly arrive, the consumer must be informed and truly understand both their options and what quality really is. The patient is the consumer, and they deserve to know what they are buying. The New Yorker claims that only 1 in 7 Americans understand the basics of healthcare plans. How can the patient demand the best quality when they don't understand the system? There are several steps that can help facilitate the change to a quality driven healthcare system. First, consumers must be convinced that there are quality problems in medical care, and that these problems can be improved (ie… I had no idea MRI diagnoses could vary so much, and in NYC alone.) Secondly, quality reporting must be standardized and universal, and the reporting must be relevant and easy for the consumer to understand, as well as widely disseminated. Lastly, insurance providers must reward quality improvements to healthcare providers.  

​It won't be an easy, but we must do our part to drive our healthcare system to one driven by quality. 

Sources
1. Spreemo 
2. The New Yorker: The Healthcare Industry's Relationship Problems 
3. Business Insurance: Higher doctor fees may get injured workers back to work faster
4. WSJ: Picking a health plan? An algorithm could help  
5. Pando: How to find a doctor that doesn't suck? Spreemo has a new answer. 
6. Commentary: Healthcare commodity, not entitlement 
7.Health Affairs: Consumers and quality-driven healthcare: a call to action 
8. Real Time: Bill Maher and Bernie Sanders debate socialism: US is 'already a socialist country'
0 Comments

Neurologic Treatments: Stalled, but gearing up

11/12/2015

0 Comments

 
Last month I attended a presentation by a biological therapeutics Pfizer VP. While several non-scientific nuances of this talk piqued me (in particular, how 30 years ago it was relatively easy to carve out a business role for yourself if you had medical knowledge), it was the focus on neurological science that really interested me. This executive said that many large pharmaceutical companies have exited neurological drug research and development. Why? The fruitlessness of the field. The exec contended that 90% of the mental health diseases today are treated with the same medication that they would have been treated with 30 years ago. While I cannot vouch for that statement, I agree the field hasn’t seen disruptive breakthroughs. Modern big pharma was arguably defined by neurologic medications – Valium in 1978 and Prozac in the 1990s –but stagnation in breakthrough neurological science has led to the exodus of many big drug companies from the field.  Although the aging population surely portends a surge in brain diseases and potential opportunities, big pharma giants such as GlaxoSmithKline, Bristol-Myers Squibb, and AstraZeneca have demurred.

While drug development has been stagnant, the research has not: I personally know scientists in Alzheimer’s whose research has shown very promising results, and moreover, there has been a spur of Alzheimer’s developments in the news lately. Nature recently published a fascinating publication from The National Hospital for Neurology and Neurosurgery in London suggesting Alzheimer’s disease could actually be transmissible. The tell-tale sign of Alzheimer’s - amyloid Beta pathology – was found in the brains of six out of eight people who had died of Creutzfeldt-Jakob disease. The shocking connection is that years before, the patients had been injected with contaminated growth hormone, suggesting that sticky amyloid-Beta seeds not only traveled along with the hormone, but actually prompted the development of disease plaques. It is believed that misfolding of amyloid-beta makes the peptide sticky, so that it forms clumps. Paired with this recent finding, the concern is that Alzheimer’s could be passed on by other routes such as blood transfusions or contaminated surgical instruments. While this research implies amyloid-plaque formation can be triggered by transmission of a seed, what of the other Alzheimer’s patients who have no history of such injections or surgeries?

A similar cadaver study implicated disparate findings, the results recently published in Scientific Reports. A lab led by Dr. Carrasco at the Autonomous University of Madrid examined the brain tissue of 25 cadavers. 14 of the subjects had Alzheimer’s when they were alive, while the other 11 were Alzheimer’s free. Surprisingly, the brains of all 14 of the Alzheimer’s patients had fungal cells in some of the neurons, while fungus was absent from the Alzheimer-free brains. While a sample size of 25 is quite small, the statistical significance of these results is nonetheless decisive. Does this mean fungi usher in the disease? Alzheimer’s progresses slowly and causes immune responses such as inflammation, and untreated fungal infections progress slowly and can trigger inflammation and blood vessel damage, which has also been observed in many Alzheimer’s patients. If fungi is responsible, this is welcome news: many medications have anti-fungal properties that could potentially be exploited to treat Alzheimer’s.  However, this study highlights a common conundrum is science: are the findings causative or associative? In other words, is it the Alzheimer’s disease that is allowing fungus to take a foothold? Perhaps there is not an ultimate cause for Alzheimer’s, but instead Alzheimer’s is a general response arising from a variety of neurological insults.  
Picture
PET(positron emission tomography) scan of three different brains. Left: Alzheimer's patient with amyloid plaques, which are colored in orange and yellow. Middle: a person with amyloid plaques but not showing cognitive signs of the disease. Right: a brain free of amyloid plaques. ​Lifelong brain-stimulating habits linked to lower Alzheimer’s protein levels
And herein lies one of the complexities of neuroscience: the intricacies of the brain and the diseases that affect it mean there is rarely just one cause of a disease.  For example, a single genetic mutation called 22q11 greatly increases the odds of an individual developing schizophrenia. However, even without that mutation, a series of small, disparate mutations can also lead to a high propensity of developing the disease. To even further complicate matters, the same small mutations that can lead to schizophrenia can lead to autism, attention deficit hyperactive disorder, or bipolar disorder in others.

Arguably, breakthroughs in neuroscience haven’t followed a deep understanding of causation. Instead,  many of the field’s most important treatments have been discovered by accident. Thorazine, the first antipsychotic, was found to stop hallucinations when it had originally been intended as a sedative. Imipramine was a failed antipsychotic but became an important antidepressant when its mood improving properties were discovered. More recent drugs such as Prozac, Celexa, Abilify, and Zoloft use the same basic mechanisms as older drugs: antidepressants boost neurotransmitters such as serotonin, and antipsychotics block the dopamine receptor D2.

However, recent scientific advancements in science could begin to convert to translational results in the mental health field. Advanced techniques such as genetic sequencing and DNA editing techniques could soon revolutionalize the field. And what that means is big drug companies are getting back in. Johnson & Johnson, Roche, and Novartis are reinvigorating their efforts into neurologics, and Forbes reports that last year, $3.3 billion was invested into companies that are developing treatments for the brain, a figure that dominates any in the past 10 years. And while treatments for Alzheimer’s and schizophrenia may be more elusive, developments in the treatment of depression, post-traumatic stress disorder, multiple sclerosis, and Parkinson’s Disease may be forthcoming. While I’d love to wax on, read here for more detail. While Pfizer may conjure up disillusionment in some, I do commend the effort to push developments in neurological treatments. 
Sources: (because if you cite a "scientific" development you should be able to show a credible scientific journal, people.... this bacon causes cancer nonsense is particularly irksome to me):
1. Forbes: The Coming Boom in Brain Medicines
2. Nature Letter: Evidence for human transmission of amyloid-beta pathology and cerebral amyloid angiopathy
Nature News: Autopsies reveal signs of Alzheimer's in growth hormone patients
3. Nature Scientific Reports: Different brain regions are affected with fungi in Alzheimer's Disease
4. The Economist: Fungi, the bogey man 
5. Bacon Causes Cancer? Sort of. Not really. Ish. 
0 Comments

//Most popular languages of 2015

11/6/2015

0 Comments

 
Picture
So many programming languages to learn! This puts it in perspective for how popular each is. I'm learning Python (for webscraping, automating analysis, modeling), R (statistics and modeling), and Javascript (D3 - for data visualization). I have done some C++. Perl, and Matlab, and I think it's pretty great to break down problems and think like a computer. 

Here's another cool infographic showing the"modern" languages- PHP, Python, and Ruby- with the purpose, usability, ease of learning, and sites built with each.  

Picture
Sources:
http://i.imgur.com/7L5ECS9.jpg
https://blog.udemy.com/modern-language-wars/

0 Comments

2006 Wall Street Trader vs 2015 Trader

9/21/2015

0 Comments

 
This article in the WSJ was kind of shocking to me: because rates haven't been raised since June of 2006, many prominent traders have not seen a rate hike in the entire tenure of their career. That is not normal. Because keeping interest rates at near zero is not normal! Yes, there is global economic turmoil, but doesn't nearly interest free debt lead to misplaced capital? Isn't that one of the primary reasons we ENTERED a recession? 

But I digress. I wanted to keep it light this Monday night. I simply wanted to point out that it amuses me how much traders have changed in the last 9 years. A decade ago, we had your Wolf of Wall Streetesque trader: Suit, tie, Blackberry, credit (ah),  mortgage back securities (AHHH). Today, we have iPhone touting, Citibike riding, exchange-traded funds traders whose risks namely include interest rates and liquidity and who  operate under the auspices of federal regulation. I gotta say, I like the guy on the right better. But are traders today more grounded to reality and more closely tethered to prudent risk taking bounds than traders of the past? Sure, maybe. Or is it simply draconian oversight, grossly absent in the leadup to the 2009 recession, that has prevented people from taking what they can get by any falsifying, embezzling, speculating means possible? 

Nothing against traders. I wanted to be one myself. It's just interesting to muse on the stereotypes. 
Picture
http://www.wsj.com/articles/ahead-of-the-fed-young-bankers-wrestle-with-a-novel-notion-interest-rates-that-rise-1442395980?mod=e2tw
Source: Ahead of the Fed, Young Bankers Wrestle With a Novel Notion: Interest Rates That Rise
0 Comments
<<Previous
Forward>>
    Picture

    Vanessa Mahoney,  PHD

    Biomedical scientist & data analyst who loves learning how things work - from mortgage-backed securities to cardiac electrophysiology to Donald Trump's comb over

     
    The postings on this site are my own and don't necessarily represent IBM's positions, strategies, or opinions. 

    Archives

    December 2018
    November 2018
    June 2018
    December 2017
    June 2017
    April 2017
    September 2016
    July 2016
    June 2016
    May 2016
    February 2016
    January 2016
    November 2015
    September 2015
    August 2015
    June 2015
    May 2015

    Categories

    All
    Biotech/Healthcare

    RSS Feed

    Categories

    All
    Biotech/Healthcare

Powered by Create your own unique website with customizable templates.