Accurate reporting of science in the media

Today, the Institute for Media and Public Trust at California State University, Fresno is convening a group of experts (journalists and a scientist) to discuss science and environmental reporting. The keynote, by Pulitzer Prize winner Deborah Blum, is entitled, "Science Journalism in the Age of Mistrust."

I am a Biology professor, with extensive training in reading and analyzing data (hence the title of this blog - Objective Compulsive Order). One of the most important things I can do to advance society is to help as many people as I can practice evidence-based decision making and critical thinking skills. Generally, in this context, these buzzwords can be translated to understanding how to:

• ignore hype 
• be a skeptic (if it sounds too good to be true, it probably is), and 
• identify (and use) quality information to make rational decisions

Because of the societal importance of this effort, I've asked my graduate course in Molecular Biology to join me this evening in the audience. I hope they'll not only learn some important things about science, but also how - as scientists themselves - they can help partner with reporters to ensure balanced and accurate dissemination of the truths these students will uncover as scientists.

Extent of the Problem of Misleading Science Reporting

To prepare my class for this event, I asked each student to send me information on a case study in which a scientific report has been potentially miscommunicated to the public through the news media. My goal was for them to explore the potential breadth of this issue, and also to be prepared to ask some questions of the panelists at tonight's event. We are not by any means the first people who have noted the sometimes unaligned interests of scientists and the media. For example, here is one analysis (itself probably more provocatively titled than it should be…): Why Most Biomedical Findings Echoed by Newspapers Turn Out to be False: The Case of Attention Deficit Hyperactivity Disorder

Below, I'm sharing a selection of my students' case study submissions, as examples of how media coverage can misrepresent research findings. The hyperlinks below lead first to the original scientific report and then to the media coverage. The link text is the title of each article.

Maybe Gru and the minions had it right…

• AP39, a mitochondrially targeted hydrogen sulfide donor, stimulates cellular bioenergetics, exerts cytoprotective effects and protects against the loss of mitochondrial DNA integrity in oxidatively stressed endothelial cells in vitro
• Sniffing your partners’ farts could help ward off disease

Fountain of youth and vampirism?

• Facing up to the global challenges of ageing
• Young blood could be the secret to long-lasting health: study

When less (words) is not necessarily more (accurate)

• Genomic Analysis of Hospital Plumbing Reveals Diverse Reservoir of Bacterial Plasmids Conferring Carbapenem Resistance
• Hospital plumbing a 'vast, resilient reservoir' of superbugs

Pro-tip: "cause" is not a word scientists use much, because it is difficult to obtain irrefutable proof of cause-and-effect (for example, read: Baseball and jet lag: Correlation does not imply causation)

• Breast size, handedness and breast cancer risk
• Bras Shown to Cause Cancer

Counterpoint: here, the media report does a great job at debunking why the science itself is probably flawed

• Women Are More Likely to Wear Red or Pink at Peak Fertility
• Too Good to Be True: Statistics may say that women wear red when they’re fertile … but you can’t always trust statistics

Hopefully at least one point becomes immediately obvious. The headlines themselves, which are meant to attract reader attention, tend to oversimplify (and sometimes exaggerate) the more nuanced scientific data. It is important to note that it is not just reporters but scientists themselves who are responsible for communicating their data. Although it is easy to conclude that journalists are trying to hype their stories so that their outlet earns money, scientists are also responsible for accurate communication as well. Both are faced with the same scenario: the goal of quickly grabbing the audience's attention. This necessitates summarizing, which always eliminates subtleties (that are often the most critical parts of scientific experiments to know about before making broad, sweeping claims). Ultimately, dear reader, the most simple and effective take-home strategy is: don't just read the headline (or media summary) and think that's all there is to the story.

Resources

There are lots of good resources on the web to help train yourself how to spot scientific data that might not be credible and to spot over-hyped or misleading news reports of scientific studies. Below are a few, curated by me. But, please, like LeVar Burton said on Reading Rainbow, "You don't have to take my word for it!"

Don't believe everything you read. Practice critical thinking and learn how to spot tell-tale signs that something's not quite right. Please, be a skeptic.

• 10 Questions To Distinguish Real From Fake Science
• Twenty tips for interpreting scientific claims
• Scientific Studies: Last Week Tonight with John Oliver (this is a really funny and fantastic video: John Oliver discusses how and why media outlets so often report untrue or incomplete information as science)
• and finally, a great web resource full of examples and tutorials on spotting fake or misleading data, run by scientists at University of Washington: Calling Bull: Data Reasoning in a Digital World

How scientists do Thanksgiving

To the objective compulsive, Thanksgiving offers great opportunities for exercising our analytical skills. I get to plan out pot and pan usage while selecting a suite of recipes that complement each other in terms of when they all have to be started to finish at the same time but not require more than four range burners and one oven with two racks, fit in the oven at the same time as a turkey and roasting pan, be compatible with the temperatures required of other oven-cooked dishes, etc.

The biggest concern for the anal-retentive is making sure all of the dishes are finished at the same time, and at the time that dinner is intended to be served! Sure, I can make use of a bevy of other techniques like keep food warm in the oven warming drawer (but I never do), and I do boil water in my tea kettle to fill the serving dishes in advance to heat them before adding the food (remember to dump the water first!) But, it is so much more satisfying to have the dishes all finish at the same time, as planned!

The master of ceremonies for Thanksgiving dishes is the turkey, of course. The turkey dictates the cooking schedule and presents the biggest challenge: knowing how long it will take your turkey to cook. If you can accurately estimate this, then it is relatively simple to work backward from the finish time to know when to start the rest of the dishes. Buy yourself a probe thermometer, and this dilemma is solved.

I take my turkey tips from Alton Brown and have been using his roast turkey recipe for over a decade, well before I met him at a book-signing ca. 2005:

Alton's recipe involves starting the brined, patted-dry, oiled turkey on a rack in a roasting pan in a hot oven (he says 500°F, I do 450°C) for a half-hour before dropping the oven temperature to 350°C until finished. The only other (slight) deviation from his plan is that I keep the bottom of the roasting pan covered in a thin layer of water to keep the drippings from burning during roasting. Burnt drippings = horrible tasting gravy later, among other problems. Other benefits: no cooked-on turkey goodness to scrub off of the pan during clean-up, and you produce some turkey stock during the roasting.

A critical component of the process is inserting a probe thermometer into the thickest part of the turkey breast when you introduce Tom to the oven. This is the best way to monitor the cooking of the centerpiece of your meal. Record the temperature at regular intervals, and you can produce the following data set (this year from a 15.4 lb. bird):

Yesterday (as with last Christmas), I was able to fit the increase in turkey temp very nicely (check out that r-square value!) to a 3rd-degree polynomial! It is useful to note that internal temperature has a predictable increase. By two hours into cooking (120 minutes), I had a really good idea that I would be pulling the turkey from the oven (always at 161°F internal temperature) two more hours hence. Let the turkey rest after it comes out of the oven - I plan on 45 minutes between removing from the oven and slicing, and my probe thermometer has revealed that the turkey breast continues to increase in temperature beyond 165°F (the target temp. for properly cooked turkey) in that period of time. I always buy one of the aluminum foil roasting pans for this resting phase. It is a great way to collect any juices that make their way out of the bird and collect them for gravy, and the high walls and lips of the pan make it easy to cover with a sheet of heavy-duty aluminum foil (to keep the heat in).

Now that I can predict turkey finish time up to two hours before oven removal, I have 2.5+ hours of lead time before serving to get the rest of the dishes prepped and cooked. This Thanksgiving, this meant:
T-70 m.: two 9x13 glass baking dishes preheat in oven with 2 Tbsp. crisco each
T-65 m.: par-boiled baking potatoes, cut into chunks, go into baking dishes for roast potatoes
T-60 m.: start boiling water and butter for stuffing (used the Trader Joe's cornbread stuffing mix)
T-45 m.: turkey out of the oven at 161°C internal breast temperature
T-35 m.: water for carrots on to boil; stuffing goes in the oven; roasting potatoes stirred
T-30 m.: defat turkey drippings and add to saucepan with 2 bay leaves to reduce to 2 c.
T-20 m.: put carrots on to gently braise; prep roux for gravy
T-10 m.: whisk reduced drippings into roux and bring to gentle boil (gravy!)
T-0 m.: everything is ready to pull, serve, and devour!

After dinner, there are two chores left: cleaning (which I don't mind) and disassembling the turkey. Last night, we only ate one breast, so we have plenty of leftovers this year (as intended). The other breast gets removed whole and refrigerated for later slicing for sandwiches. I pick over the rest of the turkey and chunk/shred for leftover dishes (typically Pampered Chef's turkey wreath recipe, my MIL's turkey curry, and turkey noodle soup). The giblets, neck, and turkey bones go in my largest stock pot, full to the brim with water, and cook overnight at the gentlest simmer (small bubbles lazily, but regularly, breaking the surface) to produce some fabulous turkey stock for the curry and the soup. This always makes extra stock that I freeze to use throughout the year in risottos and other dishes that require really excellent stock.

Salmon origami - a new butterflying (and butter-frying) approach

I enjoy eating fish, but have rarely (that is, infrequently) cooked seafood. In part, I suspected that it might not fare well with my two junior diners. However, about a month ago I bought some salmon filets and decided to pan fry them. It was a hit with the whole family, so I have made this dish just about every other week. It is great served with some rice pilaf and a vegetable.

The recipe starts with three fresh deboned salmon filets (about 0.5 lb each), 

which are liberally salted (kosher) and peppered (fresh ground).

This is where I have deviated from other recipes. One issue I encountered with cooking the filets that I'm buying is that the thickness of the cut varies from a fraction of an inch on the left side to around 1.5 inches on the right. Thus, I decided that some simple butterflying to increase surface area would help cook these filets evenly. Make a single cut as shown, but do not cut through the skin underneath the filet. The intact skin, as you will see, helps hold the two pieces together in increase stability in the skillet.

Add two Tbsp of butter to a nonstick skillet and heat over med-high heat until the butter begins to foam. Add 1-2 Tbsp of vegetable oil to limit the butter browning, and then add each filet, flesh side down (the pan is a bit overcrowded - you want to leave an inch around each filet for even browning).

After about 3-4 minutes of pan-frying, it is time to turn the fish. It would be nice to be able to turn each 90 degrees to balance each filet on one side, but they're so narrow relative to the width of the filet that they normally would just tip over. Instead, fold each filet in half so that the skin side folds together (below). This allows frying what was originally the inside of the filet to allow even cooking.

The inside of one butterflied filet now frying in butter.

Notice how pink the middle of this filet was before folding. Now this side is going to be placed face-down in the skillet. Cook each side of each folded filet until golden-brown.

Three beautiful filets finishing frying.

Unfolding the salmon back into their original filets reveals fish that is as delicious to the eye as it is to the mouth! The meat is now easy to separate from the skin using a spatula or other serving utinsel. I squeeze a little lemon juice over each piece and serve.

Leftovers (if they exist) get combined cold with baby spinach leaves, dried cranberries, sunflower seeds, and maybe some blue cheese crumbles and grape tomatoes for a lunch salmon salad the following day.

My best slide: Sex-Linked Traits

In genetics class yesterday, we began discussing sex-linked traits - like color-blindness. I'm a deuteranope (red-green color-blind), a heritable genetic disorder. I explained to students (in case they were unsure) that I'm a male (with an X chromosome inherited from mom, and a Y chromosome from dad). However, neither of my sisters (XX), nor my mother (XX) or father (XY) are color-blind. Why is this?

Human color vision is due to three opsin genes that encode proteins expressed in our eyes that absorb light of different wavelengths (generally called red, blue, and green opsins). The blue opsin gene is located on an autosome; red and green opsins are found on the X chromosome. Upon absorbance of light, opsin proteins then transmit signals to our brains to be interpreted. DNA mutations that cause amino acid changes in specific regions of opsin proteins can alter the specific colors absorbed by the opsin proteins, and thus fine-tune the colors of light that we can perceive.

In order to understand why color-blindness exhibits a sex-linked pattern (males have a much higher frequency of this disorder than females), we need to know something more about genetics. Because humans are diploid (have two copies of almost every chromosome), there are two copies per cell of the vast majority of genes. The benefit is that if one gene copy sustains a mutation (-), we still have one "good" (wild-type, +) copy of the gene to help our cells function. This tends to be true of color vision as well: if a retinal (eye) cell has one good copy and one mutant copy (+/-) of an opsin gene, the good version produces protein that allows "normal" color vision.

Because we have a "backup copy" of most genes, many mutations experienced by our genes are recessive: when there is still one good copy of the gene (+/-), we don't notice an effect of the mutation. Only when both copies of the gene have the mutation (-/-) do we exhibit the mutant trait. Under the (hopefully valid) assumption that very little inbreeding is occurring, the chance of inheriting the same mutation from both your mother and your father is very rare, which is why blue color-blindness is extremely rare in humans. But red-green color-blindness is fairly common in some human populations (and I'll address why this is important to scientists in a future post). This is entirely because of where the blue, red, and green opsin genes are found.

In order to understand why mutations on X-linked genes tend to cause higher incidence of mutant traits in men than women, we need to explore the history of the human X and Y chromosomes. Ages ago, our X and Y chromosomes were an autosomal pair and as genetically identical as any other pair of chromosomes. Then, a gene (called SRY) evolved to dictate which individuals would be male and which female. The rules here are simple: if you inherit SRY, you become male; if you don't have SRY, you become female. When SRY first came into existence, it was located on one member of an autosome pair (on the Y); the other member of the pair was thus defined as an X (lacking SRY).

Now the X and Y are (slightly) more genetically different: one has a gene that the other doesn't. This set off an amazing series of events that eventually caused the human Y chromosome to "degenerate." The human Y chromosome is physically much smaller than the X, and contains only a tiny fraction of the genes it used to share in common with the X - way back when they were still an autosome pair. The speciality field of sex chromosome evolution seeks to understand how this degenerative process occurs; this was the basis for my doctoral dissertation (but in stickleback fish, not humans).

Remember I said that we "have two copies of almost every chromosome" and "have a 'backup copy' of most genes." Human males (XY) do not have two functional copies of every chromosome: we have an XY "pair," but the X and Y are essentially genetically unrelated at this point in time. Our Y chromosome does not contain, for example, a red opsin gene or a green opsin gene. There are two tiny regions of DNA sequence similarity between the X and Y: the two "pseudoautosomal" regions (PAR1 and PAR2) that facilitate X and Y pairing at meiosis.

My best slides from this lecture (judging by student reaction) illustrate a very useful analogy for sex chromosomes:

This is entirely the basis for sex-linked traits: when a X-linked gene, like red opsin, is mutated, a mother can pass that mutant X chromosome to both her sons and daughters. However, the daughters (by virtue of also receiving an X from dad) don't exhibit the recessive trait. On the other hand, I received the Y chromosome from my father, by definition, and so my X from my mother. That X carries an opsin mutation, and I don't have a "normal" red opsin gene on my Y to provide me with "normal" color vision.

So, where did my mother get her mutant X chromosome from? Her color-blind father (my maternal grandfather). Hence, a typical pattern for X-linked traits: passed from affected grandfathers through their daughters to half of their grandsons. Why half? Because my mother has a wild-type X chromosome as well. I had a 50:50 chance of inheriting her "good" X, but I lost the coin toss.

Mendel the Magician

It is the start of the term in Biology 102 (Genetics), and thoughts turn to Mendel and his peas. Yes, I saw it in the faces of the students, and I told them that I saw it: the look. I know it because I made that face sometime during my undergraduate days. I don't recall whether it was in molecular biology or genetics, but when the instructor mentioned "Lac operon" I almost lost it. Seriously? The third time in as many years you're going to spend a couple of class sessions on the Lac operon?
So, I watch the faces of students particularly carefully when I say certain phrases, like "Mendel's peas", to judge whether they've heard this all before. Of course, my version has more detail, but in a class of 80 students, when you have to make sure that everybody is up to speed (the chemistry majors, the lone English major, etc.), it is incumbent on the instructor to try to make it enjoyable for everybody.
I thought I'd try a dramatic twist on Mendel's first law: segregation (of alleles into gametes). Mendel's laws were all inferred by analysis of phenotypic ratios of different traits in the offspring of crosses. In one cross, Mendel analyzed how the pea color phenotype (green or yellow) was inherited. Crossing green and yellow parents, he found that the F1 generation offspring all had yellow peas. Amazingly, if he mated these F1 together, the green color reappeared (as if by magic) in the next (F2) generation.

This was the teaser: how did one of the parental phenotypes disappear, and how did it reappear later? The answer is dominance, and to understand how this works, we need Mendel's first law: segregation (of alleles into gametes). Mendel had inferred that each pea has two versions (alleles) of the pea color gene, and that 50% of a plant's gametes contain one version and the other 50% the other version. The yellow parent has two copies of the Y allele (YY) and the green parent has two y (yy). So, all of the gametes from the yellow parent contains a Y; all of the gametes from the green parent contain a y.
Each F1 offspring, comprising genetic material inherited from both parents, have one Y and one y (Yy). That the F1 plants are all yellow defines the Y allele as conferring a dominant phenotype: being YY or Yy makes you yellow; only when you have no "dominant" alleles (yy) is the green phenotype evident. Assuming that gametes fuse randomly (without regard to what alleles they carry) at fertilization, we can predict the ratio of F2 generation phenotypes (yellow:green) with some simple probability calculations.
Probability? This time I saw some eye rolls and some seat-slumping. But I had anticipated and was ready for this. I handed a sealed envelope to a student sitting in one of the front rows, and then instructed each student in the class to choose a letter between A and F and write that letter down in their notes. Then I asked them to do the same: choose a whole number between 1 and 3 and write that down. "Please raise your hand if you chose the letter 'A.'" In my class of ~80 students (I hadn't taken roll so didn't know exactly how many were in attendance that day, and certainly wasn't going to take the time to count), about 9 had picked 'A.'
If you already know where I'm going with this, you probably feel the same way I did. The idea is that in a class of 80 students, if you give them a choice of six options (A, B, C, D, E, or F), then one-sixth of the students (on average) should choose each letter. 13 students should have picked 'A,' but of course the result (9) is just one sample from a probability distribution. Someday soon, when we start statistics (say, chi-square analysis), I'll use this example! Meantime, though, the demonstration isn't over yet.
"Please raise your hand if you picked the letter F." About twenty hands go up. "Please keep your hand raised if you also picked the number 2." Most of the hands drop. Four students in the class have chosen F2. "Please look around at your classmates and see that there are four students who have chosen the letter F and the number 2." Then I say, "Please open the envelope I handed you earlier and read the message inside." She reads: "Four students in class have chosen 'F2.'" Bam! Not a dry eye in the place.
And then, this!

Of course, the calculation is straightforward, I tell the students. With about 80 students in the class, about 1/6 should pick F, and about 1/3 should pick 2. Since these choices are independent events, we multiply the probabilities together to find that about 1/18 students should pick F2. 80 students * 1/18 students ~= 4 students that should pick F2.
Mendel had done the same calculations, presumably, I say. In the F1 generation of peas, 50% of gametes have Y, and 50% have y. If we assume that the gametes fuse randomly and that the two gametes that form an F2 plant are independent draws from the pool of F1 gametes, then the number of YY plants should be (50%)*(50%) = 25% of F2 plants. The number of yy plants should be (50%)*(50%) = 25% of F2 plants. The other 50% of the F2 plants are the heterozygotes (Yy and yY). Since only the yy plants have green peas (25% of the F2 generation), the rest of the F2s (75%) are yellow. Hence, there is a 3:1 ratio, which Mendel observed, of yellow:green peas.
"This was the first time I've tried this demonstration," I told the class. "And likely the last. Because if I know anything about probability, this is the only time I'll ever do this demonstration when it actually works they way I intended!"