Nonononono--never ever ever ever do this unless you are completely lost with the science itself! (It is easy to be lost with the science, but you have to be aware just how massively you're missing the point of even looking at a paper.)
You say why right there: you're not checking what the authors did. You're reading the story they want to tell!
Now, they did manage to get the story past some reviewers, which means it's probably not entirely contradicted by what the data says. But if the authors have any agenda whatsoever, if they have any academic lineage that suggests that they think about things in one way or another, if there are any financial incentives that might lead to bias, anything else--this is where they all come into play, where the authors get to explain away their own data to tell the story that they want to. (They won't think they're "explaining it away". They'll think they're "putting it in context" or "focusing on what is relevant" or somesuch. That's how bias works. It's not obvious to the person who is biased.)
You should be very suspicious of discussion sections.
Instead, you should look at the figures (and/or tables). Read enough of the figure legend, results, and/or methods so you either can understand what the figure is showing, or realize that you can't. What are they comparing to what? What were the conditions?
This is the meat of the paper. This is the distillation of science into a dense, maximally-digestable form. Here you get it as pure as it comes, ideas being tested by experiment.
Now, the meat is often really hard to chew. The discussion, where the authors narrate how the meat tastes to them, is usually a lot more approachable.
Let's take your article, Sam, as an example of how this works. You say, "We observed increased gene expression by numerous bacterial pathogens, primarily Campylobacter and Helicobacter."
Well, that sounds bad! Numerous bacterial pathogens! Although you do note--though honestly you probably could have gotten it past review if you hadn't--"Interestingly, our data suggests that these organisms are present in both ICD and control samples, but their transcripts are more abundant in ICD."
Okay, but now let's look at the meat of the paper--your data. In Figure 2, we can see the transcript levels from pathogens in every single macaque. This is awesome! We can see, in a matter of a minute or so, that (1) the distributions overlap, (2) the pathogens are about 2-5x higher on average in ICD animals, and (3) we're not sure whether you neglected to show other pathogens because every single one you show is significant increase. This is interesting, but not quite as striking as it might sound from the discussion alone, and you didn't push the discussion to focus here as much as you could have (but someone else could). Then we might think "okay, but...what about the baseline? What is the expectation?" So we go looking for a baseline and we find in Figure 2A that the Shannon diversity has gone down. If the diversity goes down, but a pathogen remains, we'd kind of expect that it'd get a larger share? Hmmm. We might note in Figure 3 that Lactobacillus genera seem to have gone WAY up proportionally (though not uniformly). We also can see in Figure 1C, and discussed in the results, that the heterogeneity of ICD animals is up. Is that important? Well, not to you--you didn't write about it in the discussion.
Now, what does one do when the figures are impenetrable and/or the importance is incomprehensible (like...what is the relevance of elevated GCSFR-2? How do we know whether it matters that it's up in one subset of the ICD animals?)?
Well, if it's impenetrable: Abstract. Discussion.
But the thing to recognize is that once you have to do that you've taken a huge step backwards to being at the mercy of the authors' own ideas. At that point you're really not doing much better than reading a well-done popular account, a review article, or something else like that. If you can't at least check that the claims made in the discussion reflect data in the figures and/or results, the value of reading the paper is only in that you got someone knowledgeable to explain to you how they feel about it. And, you know, that's more than nothing.
But it isn't what makes science tick.
So, if you can, if you care, go for the meat. That's the lowest bar to pass where having a primary research paper in your hands is substantively better than having any other type of account of the research. If you can get deeper to understand whether there are factors that the authors might have been overlooking (e.g. if ICD animals are housed in a different facility than control animals--you have to go read the methods for that one, where you find that the ICD and control animals were pair-matched in conditions as well as they could be--good! whew!--but the ICD and control animals were pair-housed with the same type...hm...).
Now, if you read enough discussion sections, eventually you might learn enough about a topic so the figures do start making sense to you. ("Oh, I remember those guys were really excited about mucin-degrading enzymes, too!")
But the figures are where it's at, if you can get there.