Direct evidence is not evidence that is "objective", or "objectively proves a fact", and it has nothing to do with the credibility, reliability or the witness's motivations or feelings. The distinction is purely based on whether the evidence requires the judge or jury to make an inference about what the evidence means. That is, whether they need to "connect the dots".
The classic example of direct evidence, in any legal dictionary, is always eye-witness testimony. For example, when a witness says "I saw John raise the gun and shoot Bill," that is direct evidence because it directly supports the assertion that John shot Bill. There are no dots to be connected.
However, whether the evidence is reliable or credible is a different issue. For example, if the defense can show that the witness was drunk, or has poor eyesight and was not wearing his glasses, or was not even present, then the evidence may be in doubt. It is still direct evidence, though.
On the other hand, if the witness says, "John and Bill went into another room. I heard a loud bang, and when I went into that room, Bill was on the floor, shot dead, and John was holding a gun with smoke coming out of the barrel," that is not direct evidence. It is circumstantial -- because now there are plenty of dots to be connected; multiple inferences are possible. Did John simply murder Bill? Or was it an accident? Or was John possibly acting in self-defense?
So what about scientific evidence then? The true answer is that it depends on what the scientific evidence (DNA, fingerprints, etc) is being used for. If the scientific evidence is being used to support the assertion of guilt or innocence, then it is almost always circumstantial -- because it is requires the jury to connect the dots.
For example, the defendant's bloody fingerprint is found on the murder weapon. Does this directly prove the defendant's guilt? Why, no! There could be many explanations for that. What the evidence does show is that the defendant handled the knife after the knife became covered in blood -- but whether that means the defendant killed the victim will depend on, well, circumstances. For instance, if it can be shown that both the defendant and the victim were alive when they entered the scene, that no-one else could possibly have been present, and that the wounds could not have been self-inflicted or have occurred by accident (that is, the victim did not fall on the knife), then the fingerprint is highly significant -- but it is still circumstantial, because inferences have to be made, and those inferences will depend on other factors -- that is, the circumstances.
The same is true of DNA or any scientific evidence. DNA evidence is not cast-iron proof of guilt, and it does require inferences to be made and circumstances to be taken into account. The presence of DNA may prove someone was present, but it does not, for example, prove when they were present or what actually happened. In the case of sexual offenses, for example, DNA may prove contact -- but it tells us nothing about when that happened, or whether there was consent. That is why it's circumstantial, whereas the evidence of a witness who says "I saw John rape Mary" is direct.
In closing, I'll support what I'm saying by quoting the famous prosecutor and author, Vincent Bugliosi: "Circumstantial evidence has erroneously come to be associated in the public mind and vernacular with a weak case. ('Oh, that's just circumstantial evidence.') But nothing could be further from the truth. In fact, most first degree murder cases are based on circumstantial evidence. This is so because other than eye-witness testimony (and in some jurisdictions, a confession), which is direct evidence, all other evidence, even fingerprints and DNA, is circumstantial evidence."
I hope this helps.