I sat down the other day to prepare a figure for a paper I’m co-authoring, confident I’d have it dispatched in a matter of minutes.
You can see where this is going already, can’t you?
It was a relatively clean Western blot result – a black band on a grey background, and no other offending pixels. The sort of biochemical result that is so unambiguous and clean that you don’t bother getting more than two or three exposures when you’re in the darkroom – especially when you’ve been busy flirting with that mysterious stranger with the New Zealand accent. After I scanned the X-ray film into a tif file, I could see that it certainly wasn’t perfect – the film we’ve been using tends to come out quite dark, so I knew I’d have to fiddle with some settings in Photoshop to make it publication quality. But I didn’t really expect it to be a problem.
Pure, unadulterated biochemical genius*
A few hours later, I was stumped. To make a long story short, what looked beautiful on screen looked like a dog’s breakfast when printed out; and what looked decent on a printout looked awful on my computer. I tried Levels, I tried Curves, I tried Desaturation; I tried sacrificing a vial of virgin Drosophila to the Photoshop gods. Despite the unambiguity of the result, making the band stand out cosmetically in both formats seemed impossible. And I think it’s important for a figure to look good both ways; after all, a referee might look at your images on screen, or print them out. There is really no way to tell how your data might be consumed. And superficial appearances, unfortunately, are often more important than substance.
Photoshop is evil
This wasn’t the first time I’d hit a wall in image presentation. A few days previously, I’d had a similar problem trying to print out a rough result for my lab notebook. This had been a three-color confocal montage from the Leica SP5, processed through ImageJ, adjusted in Photoshop and mocked up in Illustrator. Heart-stoppingly beautiful images on-screen – but they were almost invisible on the printout, as if my cells were being viewed through a thick smog. Various postdocs gave me various bits of advice – adjust the Levels; convert to CYMK; pull the legs off of virgin Drosophila while reciting the alphabet backwards. I was told various bits of conflicting lore: for example, using Levels and Curves isn’t formally cheating, but adjusting the Brightness/Contrast is. And vice versa. After an entire day of this, I gave up, and the murky, uninformative printout in my notebook contains a scribbled caption that says, rather defensively, “this looks a lot better in the electronic version, but it seems that actin is decreased on over-expression of the gene”. No future archivist will ever believe it, though: the rare case of a thousand words being worth a picture.
At that point, I started to try to remember the last time I’d mocked up a data figure for publication. Why didn’t I seem to know any of the Photoshop tricks? It was only then I realized that I’d actually never done it before. When I was publishing in Leiden, I had a team of people working for me, and they did all the image processing for our papers. And before that, as a post-doc in London, none of my figures required this sort of manipulation, being mostly graphs and regular photographic images of cell cultures under phase contrast. Even more previously, as a graduate student, you took your blots and gels, mocked up with glued labels, down to the friendly guys in the Photography department and casually suggested that they might consider downplaying this or that particular background band when they were committing your data to old-fashioned 8×10 glossies.
It was at that point, with sinking heart, that I realized I might actually have to take a Photoshop course.
Oh dear God, no.
There’s a lot to be said for old-fashioned 8×10 glossies. Ah, the smell of the spray-mount and the roar of the Letraset.
I just used a printer, paper and a Glu-Stick. Does that make me common?
Sometimes simple is best.
The J Cell Biol put out some guidelines on what is/isn’t acceptable when manipulating images.
I doubt there is a Photography department left in any university on the entire planet.
Frank, that’s very useful. This line is interesting: ‘Adjustments of brightness, contrast, or color balance are acceptable if they are applied to the whole image and as long as they do not obscure or eliminate any information present in the original. Nonlinear adjustments (e.g., changes to gamma settings) must be disclosed in the figure legend.”
What’s a gamma setting, though?
I ws going to say what Frank’s link said: that you can totally adjust the brightness/contrast as long as you do it for the WHOLE image (including controls). It’s when you up the contrast in figure panel 3A but not in panel 3B (“because you can see it fine there“) that you’re in trouble.
Meanwhile, I foolishly believe I can take PhotoShop in a fist fight and I downloaded that first image to try and play with it. If it works I’ll e-mail it and tell you what I did so you can repeat it on the high-res. If it doesn’t, I’ll swear a bit and never mention it again.
Be afraid. Be very afraid. We Nature editors are increasingly on the look-out for image manipulations that stray from the ‘honest-Officer-I-was-just-trying-to-make-the-band-stand-out-a-bit-more’ variety and across the border into something murkier. There might be a lot of it about. Keep your nose clean. Move along, now, there’s nothing to see.
One day I’ll tell you the story of Latimeria photoshopensis.
There’s an interesting undercurrent there – we trust the machine’s output (or the result of whatever it was that Jenny got up to in that darkroom), but as soon as it gets into a computer setting, all bets are off.
My digicamera at home has so many settings, light adjustments, white balancing and all sorts of other things that are in the manual I’ve lost, I’m sure some of it is as advanced as some basic photoshopping.
So the line for journal editors to draw must be quite complex.
What would happen if you sketched it? Would you end up with something monochrome, and thus distorted?
So the line for journal editors to draw must be quite complex
If I told you, I’d have to kill you.
You should have set the resolution to 300dpi, Jennifer , on screen it makes no difference because all the computer screen does is represent things in pixels, whereas printing depends on how many pixels make up each inch of printed area.
Image> Image Size , set resolution to 300 pixels per inch, bicubic smoother in the drop-down box, and you’ll get sharp printouts.
Obviously I’d set the resolution at 300 – having been a journal editor myself for a number of years, I am familiar with the requirements of print resolution. My point was it didn’t help – something else strange was happening.
Eva, the image you sent me is very pretty! Can’t wait to try it at work and compare it to the efforts of another local expert. I could give a prize.
Henry, it must be hell when cheeky authors attempt to Photoshop a few extra limbs onto their exhumed skeletons just to make them look more interesting.
Henry, it must be hell when cheeky authors attempt to Photoshop a few extra limbs onto their exhumed skeletons just to make them look more interesting.
Yes, how we laughed on the way to the A&E.
Jennifer,
have you already thought about calibrating your monitor and/or printer? Usually when the printed stuff looks different than what you see on screen, it is a calibration issue. There are software tools for eyeballing the calibration, but hardware is more precise (and less subjective). And tools like the Spyder are not that expensive!
Alexander – no, I hadn’t! Although of course that won’t help when a referee with another printer prints out my images…what I’d like to achieve is a way is a quality that has a high chance of looking good no matter who is printing it, or viewing it on screen. Maybe it’s not possible.
Colorsync profiles used to be the way of doing this: but they depend on everyone using them.
I was looking a bit more carefully at that link Frank sent. Some of the instructions said it is wrong to Photoshop a blot so that a background band goes away. But then, this is exactly the same thing we do in spirit when we do a shorter exposure of the raw data. Which I think is the same thing Scott was saying with his comment we trust the machine’s output … but as soon as it gets into a computer setting, all bets are off.
So is it immoral to use a naturally generated shorter exposure when you know that a longer exposure will lead to background bands? If you have the proper controls, background bands are actually not damning in themselves – they just distract from your visual message. And of course if you leave a blot long enough on film, the whole thing just turns black. But this is reality as well.
Ah, Alexander mentioned calibration. Whole chapters of very weighty Photoshop books have been written about this, and it’s fiddly. Also, it has the side effect of ensuring that what you print is reflective of what you see on your screen, but depending what your printer’s like it might result in your computer screen displaying everything else like some horrid dog’s breakfast of evil contrast and colour.
And, of course, it will have nothing to do with any other printer you might wish to use later (like one at your friendly graphics department, that you don’t have any more).
I had a team of people working for me, and they did all the image processing for our papers
This reminds me of my dad’s stories of having the art department draw his graphs for his publications. In ink. With a pen.
Regarding gamma, you’ll find it (in Photoshop CS2 and later) in the Image/Adjustments/Exposure dialogue box. Purists might disagree, but the middle slider in your “Levels” dialogue is also gamma (or gamma-like), i.e. it makes the midtones lighter or darker.
This explanation is probably wildly inaccurate, but you get the idea.
Photoshop – big, big learning curve, but ultimately rewarding. I’d say get a good book and prepare for a lot of fiddling around. 🙂
Ah, now we’re getting philosophical. Is a Western blot ‘real’ in any sense of the word, hmm?
I blot, therefore I am. Oh God, I need a drink. I’ve just realized that the Sainsburys own brand decongestants are phenylephrine and not pseudoephedrine, which is probably why I’m still feeling like crap.
As coincidence would have it, our institute is trialling one of those machines that processes chemiluminescence-based Western blots, all this week. The machine arrived yesterday and is actually sitting in our lab space; today I overheard lots of postdocs chatting about exposures and background bands and how to make the image look its best. I guess this opens up even more room for interpretation.
‘interpretation’ is a euphemism, right?
My Boots own-brand decongestants are phenylephrine too – but they had the same active ingredients as the Beechams and other brands I checked out. Pseudoephedrine drugs are harder to get now – I think you may have to ask over the counter for them.
And Jenny, yes, what I was getting at was that all the suspicion and requirements seemed to be on what you could or could not do with the image once you’d got it out of The Machine, rather than what The Machine could do – and there must be cases where you could vary (or there would be variance) in the output from The Machine pre-Photoshopping – eg calibration or in-machine modification.
Yes Scott–over the counter at Tesco actually came up trumps. \o/
My feelings about this are tempered by the thought that all data need to be taken on faith no matter how they are generated. If someone really wanted to commit fraud, they could just load positive control sample into a lane and claim it was their experimental sample. You either trust a researcher, or you don’t. I think because the tools exist now to make things look prettier, people will feel the pressure to use them – and it’s more to do with expectations than wanting to mislead.
Mind you, I was greatly cheered by a talk I attended a few weeks back by Caroline Hill. She showed about a squillion Western blots and they were gloriously messy: curving fronts, speckles, smudges, the works. Yet despite this the results were crystal clear. I really admired the honesty behind that biochemistry. Why re-run a gel again, wasting time and resources, just to get a straight front, when the result is already unequivocal? It’s just fear of negative referee response to the cosmetic that probably induces people to want things to look perfect.
This whole “in the Machine” vs. “outside the Machine” argument is silly anyway. The machine has software in it. The computer on your desk has software in it. If the image captured by the CCD (or PMT, or whatever) inside the machine was given to you raw, it would be so ugly that even BBRC wouldn’t publish it. Software massaging is always required, so I don’t really see (a) what the difference is if you do it on your desktop/laptop/supercluster, or (b) if you fiddle your exposure times and contrast on the instrument in order to get “the best exposure”.
As somebody or other said up there ^, it’s
just the sameanalogous to different film exposures, like back in theprecambrianold days.Addendum – as long as the control lanes are treated on the same blot along with the experimental lanes, everything should be representative of the results of the experiment as a whole.
Come on, journal editors, bring it on, I dares ya… [dances around threateningly]
it would be so ugly that even BBRC wouldn’t publish it
snort
/Deletes mental image of Wintle mud-wrestling with BBRC editor.
Heh. I struggled for days to get one of my Western blot results to look even
halfquarter1% decent in my PhD thesis. I could see a beautiful band – if I waggled the autorad film up and down in front of a piece of white paper – but it just would not reproduce in print. (This was a notoriously awful antibody that needed at least a half-hour exposure to see anything). I ended up just submitting the damn thing with a nasty blurry mess, and then took the actual film to my defense to show the examiners if they commented on the figure (they did, and were most amused when I pulled the film out of my bag). I got away with it, but had to insert a disclaimer about the quality of the image in the text. Luckily it wasn’t a major point – just a confirmation that a protein that had no discernible effect on my cells was actually being expressed.If the referees happen to live somewhere lovely, like Hawaii or Barbados, I’d be happy to go show them my original films in person!
A girl in my lab once had to show the editor her original films. But she just had to mail them and then they compared them to the figure she submitted and it was fine and they mailed the film back and published the paper. (She’d cut out the middle lanes because it was another experiment, that wasn’t included in the final paper, so they needed proof that both sides of the figure came form the same original film.)
Wow, that’s editorial dedication. But again, it seems misplaced: if you really wanted to commit fraud you’d just lie about what samples you loaded into your lanes.
In more traumatic news, I just spotted the mysterious stranger with the New Zealand accent developing his blot on the fancy new demo machine. Which means I’ll be all alone for my ritual Friday afternoon darkroom session.
(You might be wondering at this point why I didn’t trial the machine myself, since I had a blot to do. Answer? Couldn’t be arsed to take the course, and the activation energy barrier is just a wee bit too high after a late night of Thanksgiving debauchery. Besides, Dr Kiwi took the course and he’s been over there swearing at the screen for hours. Albeit in a lovely accent.)
Is that the kind of machine where your blot goes directly on it and you close the door to the box and then the bands come up on the screen?
They’re not as sensitive as film (at least, the ones I’ve used took 30 minutes to see a band that our film picked up in 10 minutes – so if there’s anything longer than 10 minute exposure on film, it won’t work well, because the signal dies out) and they form huge line-ups because only one person can use them at a time. Nobody at my institute used it for exposure, but I’ve used it to quantify bands from film.
Today was the last day of the demo, and there were loads of swearing postdocs and a lot of head-scratching. Apparently this is one of the cheaper models – e.g. can’t do fluorescence.
I was so happy in the darkroom – it’s just so much easier. And I got a great result!
Ha. X-ray film, darkroom, and powerful, powerful radioisotopes. Your ticket to happy results Nirvana.
Possibly.
Why is anyone still using film and a darkroom? That’s so…old fashioned! I haven’t used film and a darkroom since grad school. Since then it’s chemiluminescence all the way! Faster, funner, toxicer!
Oh, and Image J = Satan’s Balls. Horrid bloody program.