- I immediately think, "Oh my God, I'm being framed."

(suspenseful music) "This is impossible.

I cannot explain how this could be."

Meredith's DNA should not be on the knife that was found at my boyfriend's house.

- Amanda Knox spent four years in prison for a murder that she didn't commit.

The media circus was honestly pretty insane, all because some tiny invisible molecules seemed to say that she did it.

- [Reporter] Her pleas fell on deaf ears.

Amanda Knox is now a convicted killer.

(tense music) - [Amanda] A lot of magical claims were made about how, if you find DNA, that it's unquestionable and it definitively proves guilt.

- Forensic DNA technology's become so sensitive that today, a single cell's worth of genetic material can be detected at a crime scene and identified to solve that crime, or at least that's how it's supposed to work.

But- - Just because there's DNA doesn't mean we know what happened.

(dark music) - What happens if we push that science beyond its limits?

What happens when police start chasing ghosts?

(mysterious whimsical music) Hey, smart people, Joe here.

Forensic DNA technology has totally revolutionized criminal investigation.

(mysterious music) We have the ability to collect, process, and identify invisible, beyond microscopic molecules that are left behind at crime scenes, which is honestly mind-blowing.

We're all probably somewhat aware of what DNA forensics is, and for many people, I'm betting that's thanks to TV.

♪ Yeah ♪ (energetic music) - Shows like "CSI" created this perception.

Researchers literally call it The "CSI" Effect, where the public and more importantly, people on juries, came to see DNA evidence as basically magic.

Swab a crime scene, run it through a computer, and boom, you got your killer.

Juries kind of tend to treat DNA as infallible proof rather than one piece of a much messier puzzle.

- That DNA doesn't lie.

(screen buzzes) - I think it's important and pretty interesting to learn how DNA forensics actually works.

And as we're about to see, it isn't always as infallible as we've been told.

I guess this is a true crime show today.

(low mysterious music) On a sunny spring day in 2007, two German police officers were taking a break in their car.

They had the doors open.

It was a really nice day.

But moments later, they were both shot in the head from behind.

One officer, 22-year-old Michele Kiesewetter didn't survive.

Investigators diligently swabbed the scene to collect any traces of the killer's DNA, and they got a hit.

To their shock, it matched DNA from multiple other crime scenes.

This suspect's DNA was connected to drug deals, burglaries, and assaults in multiple countries across Europe, in addition to a half-dozen other murders.

A serial killer and super criminal was haunting Europe.

The strangest part?

All the DNA evidence pointed to the fact that the killer was a she, despite the fact that only one in six serial killers are thought to be women.

She was tied to several different accomplices, none of whom would give her up to police.

Authorities figured they were all just too scared of her to snitch.

Police eventually linked her DNA to 40 different crime scenes across Europe, going back to 1993.

She was dubbed the Phantom of Heilbronn because no one ever saw her commit these crimes, but her DNA was everywhere.

But fast forward to April, 2008, as evidence is piling up, investigators announced, "We're closing in on her."

Except there was no her to close in on because the killer that they were chasing wasn't real.

Everything started to fall apart when the Phantom's DNA showed up in a French case involving a burned body.

(dark music) Now, the police had a hunch that the body belonged to a man who'd gone missing, but this person had entered the country seeking asylum, so investigators had his fingerprints on that asylum application.

But when they swabbed those fingerprints from his asylum application to serve as a reference to ID the burned body, the Phantom's DNA turned up in that sample, female DNA.

This didn't make any sense.

The swab of DNA wasn't even from a crime scene, and it should have only contained male DNA.

A horrifying realization began to hit the investigators.

They had indeed been hunting a phantom, but that phantom was DNA contamination.

- So, the DNA's strength, like all of our strengths, is also its weakness.

- [Joe] That's Dr.

Greg Hampikian, a forensic DNA expert.

- I've gotten DNA out of, you know, 400-year-old mummy bones.

It's very persistent if it's taken care of.

It lasts quite well, even at room temperature.

Every year in every lab, there are contamination events.

(mysterious music) - In the Phantom's cases, the contamination turned out to be from the cotton swabs used to collect samples at the crime scenes.

Those swabs had come from a factory in Bavaria, where they'd been contaminated with DNA from a woman who worked there at the factory.

Her DNA matched the Phantom profile at all of those crime scenes.

A few stray molecules had led police on a wild goose chase across Europe for years.

- It's very easy to move DNA, and we should never confuse identification information, which is, who is the ultimate source of this?

With questions about activity.

How did the DNA get here?

- Of course, I had to ask Amanda what she thought of this case.

- She is like, "What?

I just make cotton."

(Amanda laughs) I mean, what incredible story to talk about how, like, narratives form and, like, completely take over the public imagination, and they have a completely different explanation.

Like, "Ah, wow."

- Instead of just focusing on the identity, which is the only thing that DNA can actually tell you, the police made up a story about activity, about what had happened, what happened to get the DNA there.

This was the fatal flaw of the Phantom of Heilbronn investigation.

They invented a serial killer.

And that supposedly magical DNA evidence wasn't so magical here.

To understand how a cotton swab fooled the entire German police force, we need to understand what forensic DNA analysis actually is and why it's both incredibly powerful and incredibly fragile.

So, how's it supposed to work?

Let's solve a murder, shall we?

(mysterious music) Imagine this kitchen knife was collected at a crime scene, and it's believed to be the murder weapon.

To catch the killer, we've swabbed the knife, hoping that the murderer left behind some traces of bodily materials like skin cells, hair, blood, or semen, because all of those contain DNA.

Now, importantly, we also have a swab from the victim to make sure that any DNA that we find isn't just the dead person's DNA.

A DNA from a single skin cell contains a person's entire genome, billions of DNA letters of their full genetic blueprint, but investigators today don't usually read all of that DNA to get an ID.

Sequencing whole genomes is getting mind-blowingly cheap and maybe how forensic DNA testing is done in the future, but today, we're gonna use a different cool trick to read a genetic fingerprint that is currently the gold standard in forensics.

Our genome has thousands of genes that carry instructions for making stuff that keeps us alive, but the vast majority of our DNA does not make stuff.

And in those giant in-between regions, there are these short little stretches, where DNA sequences repeat back to back.

They're called short tandem repeats or STR.

We have lots of different STRs throughout our genome, and each STR can vary in how many repeats it contains.

The sequencing, just 20 of them, can create a DNA profile that can identify one unique individual.

To show you what I mean, let's take a look at the DNA profile of our victim.

Now, at the first STR spot, the little DNA repeating sequence is AGAT.

The victim has one allele or version of the STR with five repeats, and another allele with nine repeats.

That's because STRs are something that we inherit from mom and dad, just like all of our other DNA sequences, so it's possible to have two different alleles for any STR sequence.

- They say a-le-le in Italian, sorry, alleles in English.

- The second STR spot consists of repeating TCATs.

Our victim only has an a-le-le for 12 repeats at this STR because they inherited the same one, 12 repeats from both parents, mom and dad.

Now, the third STR spot consists of repeating ATTs, and our victim has alleles for eight and 10 repeats.

We do this on and on, up to the 20 STRs that make up a full genetic profile.

These 20 STRs are standard across forensic investigations, and they form the basis of something called CODIS, the massive DNA database maintained by the FBI.

All of the cells in the victim's body have their unique combination of STR repeats.

except they're eggs or sperm, which only have half of them.

But the most important thing, the thing that makes forensic DNA identification possible is that the victim has a different combination of STR repeats than the murderer does, and not just that.

They have a different DNA profile than any other person does.

Check this out.

So, two people may share the same number of repeats at one STR, maybe even two, but across all 20 STR sites, there is less than a one in a trillion trillion chance that they have all of the same repeats at all of the same spots, unless they're identical twins.

Those are pretty good odds that any DNA fingerprint is totally unique.

So, let's go back to our murder weapon.

The swabs will have picked up DNA from our victim, and now that we have the STR profile from our victim, we can easily separate it from any other STRs that we may find in the mix.

Any other DNA profiles we find, it might be the murderer.

We can then run those STR profiles against those of any suspects to see if there's a match, or run it through CODIS to search for matches from this giant database of criminals and other crime scenes.

And we found another DNA profile that matches someone near the case.

So, does that mean it's our murderer?

Oh, not so fast.

First, I need to tell you about Lukis Anderson.

5,600 miles away from Heilbronn in Santa Clara County, California, Lukis Anderson was having a very bad day.

In December, 2012, his DNA was found in a sample taken from under the fingernails of a wealthy California investor who'd been tied up and killed during a home break-in.

Lukis's name popped up as the match to that DNA because he was in the database due to his past criminal record.

He also had memory problems from an old head injury, and he routinely blacked out due to alcoholism, so he honestly couldn't say for sure that he was innocent.

To the prosecutors, this looked like an open-and-shut case.

The DNA analysis worked as it should.

It provided an unambiguous match to a known criminal, and there was plenty of circumstantial evidence to write a story of what happened that night.

Lukis was sent to jail on a murder charge, which could carry the death penalty.

But as his lawyers combed through evidence, trying to save him from death row, they found something pretty important.

The night of the murder, Lukis had been in the hospital for alcohol intoxication after collapsing at a store.

So, how could his DNA be at the scene of the crime when he was in a hospital bed miles away?

His defense team interviewed the paramedics who picked him up, and it turned out that they were the same paramedics who responded to the 911 call at the scene of the murder, and they had processed the victim's body.

- Human DNA can transfer as solid, as a liquid, or as an aerosol.

- Lukis's case was one of DNA transference.

Tiny traces of DNA had traveled from Lukis to the paramedics, to the victim.

But because our scientific technology is so sensitive, that was more than enough to yield a positive match to Lukis.

We don't know for sure, but his DNA may have hitched a ride on a pulse oximeter, you know, those little things they put on your finger, that the medics had used on both men.

- We call it activity level proposition.

Can you tell when it's directly by a hand?

No, there's no way to tell, but there are people in my field who will testify that way, and I think it's shocking and fraudulent.

- When do people stop asking, is there an innocent explanation for this DNA?

Are we taking shortcuts to automatically assume the guilt of the DNA and not asking that question?

- Thankfully, Lukis was eventually freed, but only after he spent five months in jail.

And the prosecutors weren't the ones asking the questions.

The science did exactly what it was supposed to.

Lukis's DNA genuinely was there, and that was enough for them to construct a narrative about what that DNA meant, a narrative that could have sent an innocent man to death row.

And this is nothing compared to what happened with Amanda Knox.

(dark music) 20-year-old Amanda was studying abroad in Perugia, Italy in the fall of 2007.

She was living with her roommate, Meredith Kercher, who just started dating a local student named Raffaele Solacito.

On November 2nd, the unthinkable happened.

Meredith's body was found in her bedroom of their shared apartment.

She'd been sexually assaulted and brutally murdered with a knife.

Now, the DNA and fingerprints of a man named Rudy Guede were found all over the crime scene.

He had fled to Germany.

He was arrested and ultimately charged with the murder.

But during the investigation, Amanda was subjected to an endless series of harsh interrogations, ultimately resulting in a coerced confession.

Amanda and her boyfriend were arrested and also charged with Meredith's murder.

- It affected the way DNA was then processed afterwards, so the arrests came before the DNA came.

I think it took two or three weeks for the DNA that they collected at the crime scene to be processed.

And by that time, they already had arrests.

They had already announced to the world that the case was closed.

They already had a narrative about what had happened to my roommate, which turned out to be very, very wrong.

I'm a 20-year-old kid who believes that when the actual evidence is processed at the crime scene, it will reveal the truth that I had nothing to do with this crime and that I'll be released and let go.

- But that's not what happened.

(suspenseful music) Investigators found a torn-off bra clasp from Meredith's room and a kitchen knife in Raffaele's apartment.

Forensic analysis identified Raffaele's DNA on the bra clasp and Amanda's DNA on the knife handle and Meredith's DNA on the knife blade.

Amanda was convicted of murder and sentenced to 26 years in prison, thanks in large part to this supposedly slam-dunk DNA evidence.

I mean, her DNA on a knife handle, a victim killed with a knife.

What else could that mean?

Well, as we've seen, DNA evidence isn't always as clear and unambiguous as TV would lead you to believe.

- You have DNA experts at war with each other who are each claiming to have authority over the truth.

(dark music) It was just incredibly frustrating for me to have to sit there and listen to people construct a reality out of whole cloth all for the sake of the narrative.

- Four years passed before the court agreed to hear Amanda and Raffaele's appeal and re-examine the evidence with a panel of independent forensic experts.

- Our crime scene was processed by forensic experts twice, once in the immediacy of the crime being discovered.

But then after those samples had been processed and none of it implicated me or my boyfriend or anybody in the crime, they went back in 46 days later to try to find new forensic evidence to implicate us in the crime, and they have completely upended our house and contaminated it completely.

- The only piece of DNA evidence tying Raffaele to the crime was found in that problematic second sweep after evidence had been mixed and thrown around, all but guaranteeing that the evidence would be contaminated.

Oh, and the knife?

- I did a little work on Amanda Knox's case.

- I love Greg!

He's the best!

- Yeah, really, it was just that knife for Amanda.

A single cell can make it look like a suspect was on a piece of evidence.

When we saw the electropherograms, it was clear to us that the DNA was at such a low level.

How are the Italians looking way below anything I do in my research lab?

- The trace of DNA that is on it is so minuscule that it can't be attributed to anyone.

- I knew right away this was suspicious.

The guidelines say you need to separate evidence and reference samples.

You do all of the evidence first.

When the evidence is finally processed, you bring the reference samples in the lab.

You never bring 'em into the lab before that.

- Reports indicated that the knife blade samples were processed in the middle of 50 samples of reference material from Meredith.

Traces of her DNA could have easily contaminated the knife blade samples.

And considering that modern forensic DNA technology is so sensitive, it can pick up on just a few cell's worth of DNA, not following careful contamination procedures is incredibly risky.

It was another case of confirmation bias.

Investigators had constructed a grizzly narrative of this crime and pushed the science well beyond its safe limits until they found data that agreed with their conclusions.

This is the opposite of good science.

- The independent experts found the exact same findings that my defense team found.

- We showed in the end it was probably transfer contamination and not the result of a criminal act that took four years of Amanda being in prison to convince the Supreme Court in Italy.

- [Amanda] I'm really interested to see a world in which the forensic evidence comes so quickly that it can be the thing that guides the narrative construction as opposed to the thing that tries to justify an already established narrative.

- In a world like the one Amanda describes, with faster, better forensic science, how many other people would never have been arrested for those crimes they didn't commit?

- Anytime you introduce a new process, you introduce a new field of error.

That's always gonna be true, no matter how good the instrument is, no matter how good the training is.

- Because science is evolving at a rapid pace, especially science around DNA, it's so important that we open up opportunities to appeal decisions that were based on bad science in the past.

- There's a couple of states that have passed laws that say, "If the science you were convicted by has now been superseded by better science that corrects the old view, you have a right to retest evidence."

Most states don't have that law.

- And so, innocent people are sitting behind bars with no legal avenue for proving their innocence.

- How many people have been convicted with sloppy DNA?

How many times has that happened?

I'll bet it's happened a lot of times.

- The beauty of science is that science it recognizes that it has its own limitations, it's an evolving practice, and it's willing to revisit and re-evaluate certain calculations.

The problem is that legal systems are not designed like that, and I think that the legal system would do well to take on a more scientific mindset of, "It is possible that we got this wrong."

- We could all use a little more scientific mindset.

Something that kept coming up in these conversations and cases was how often humans fall victim to constructing stories, and then looking for evidence that supports those stories, rather than following the evidence and building the story based only on what can be proven.

This is what science demands, right?

With technology that's now sensitive enough to get a DNA profile from a single cell, it's more important than ever to avoid weaving elaborate stories from a molecule like DNA, because DNA can really only answer the question of who, not how.

With the Phantom, Lukis, Amanda, and who knows how many other cases, the DNA never lied.

We just didn't ask the right questions.

Stay curious.

- Science is so interesting!

(laughs) - The researchers... (trills) Literally.

We can then run those STR profiles.

Profiles.

(trills) Let's go (beep).

(object slams) I meant to do that.