 And with that being said, I'd like to introduce our next speaker for this evening. So she's a medical biotechnologist with a PhD in cancer research. And she'll be talking today about extending DNA analysis. So please welcome Anna Moulner. Well, thank you for being here. And for the kind of introduction. So just a quick recap. I'm a biologist by training and did my PhD in cancer research. Now I work as a PR consultant for health care as my day job. And I'm also a science communicator. And this is seen as something of science communication has nothing to do with my day job. And I wish to talk to you about DNA. You've probably heard of it. And it stores our genetic information. So it's a different kind of code than most people here are used to. It's made up of four bases, adenine, thymine, guanine, and cytosine, which are usually abbreviated ATG and C. And as you can see in the image, the adenine always binds thymine and guanine always binds cytosine and vice versa. And so you have two negative of one positive code always. And there's coding and non-coding DNA. So the coding DNA is actually genes. So everything that shows up in your looks. And the non-coding DNA is not genes. So everything outside of your genes, both are important. And DNA you might have heard is also used as evidence in crimes. Whereas 99.9% of our DNA is the same, which is a good thing because otherwise we'd all be mutants. 0.1% is different and differs from person to person. And in 1984, Alec Jeffries discovered DNA profiling, where he saw that he could use this 0.1% to identify people. And this is possible to compare the DNA profile of one person with another person and can also be used to find relatives for paternity issues, for example. And often this is referred to as DNA fingerprints. And what this really means is checking for short tandem repeats or STRs. And these STRs are repetitive sequences outside of genes. So they have nothing to do with how we look like. And there are many STRs with different lengths, which are inherited. And the pattern of these STRs differs from person to person. And basically it just looks like this, that you have, for example, the repeat CTA five times or six times or nine times. And the number of the repeats is the thing that differs. If you look at the gene then, this could look something like this. And basically, you just count the number of the repeats. So each box here represents a repeat. Sometimes you have a similar one and the same number. And sometimes you have different numbers. What do you do to get a DNA fingerprint? Is first that you isolate the DNA from a crime scene or from a probe. You amplify these STRs. And then you can separate them by their size because the different number of repeats generates a different size of the amplified sequence. And this you do to 16 STRs. And then you get a profile that looks something like this, which you might have seen in the TV series already. And what I said was that the DNA fingerprints don't really show how you look like. However, because they are outside of genes. However, as you may remember, we have chromosomes. And two of these chromosomes are the so-called sex chromosomes. And these are called X and Y. And you can identify the X chromosomes and the Y chromosomes by the gene of the enamel, of our tooth enamel. And since 2004, we can use this to identify the sex chromosomes of a person because the X chromosome gene is six bases shorter. But this is not always accurate. So this is how the DNA fingerprint then looks like for the scientists. This is an electro-paragraph, basically just the measurements of the bands that you just saw. And you can also see on the lower left the X chromosome peak. And you see there's no Y chromosome peak. So this person has two X chromosomes. And this is just a means of identification. So it does not show much about the person besides the sex chromosomes, where you can determine the sex of the person. And in 2016, there was a rape case in Freiburg, an actually rape murder, where Maria L was found raped and murdered at the Dreisarm River in Baden-Württemberg. And after a long search, they found one hair that was found in a hedge. And it was 18 centimeters long, and it was dyed blond. And on a surveillance video of Tram, they could find a person that fit this profile of this hair style, which led to an arrest. This was a 17-year-old Afghan refugee, and he confessed that he committed the crime. And afterwards, the police said that with extended DNA analysis, the murderer could have been found much sooner. But what is this extended DNA analysis? There is now a new law to come, so not already here yet, but proposed by Baden-Württemberg, which is now governed by the Green Party, and Bavaria, which is governed by the Conservative Party, CSU. And they wish to draw more conclusions from our DNA. So they want to look at the eye color, the hair color, the skin color, and the biological age. And the Conservative Party also wants to look at the biogeographical ancestry that can be determined from the DNA. And this is basically just a synonym for ethnicity, and some may even say for race. And the law for extended DNA analysis is often referred to as the DNA facial composite. And however, in science terms, it's called forensic DNA phenotyping. And this means that you look at genes which determine the looks or other characteristics of a person just from the DNA and the genes. So what you do here is that you check for single nucleotide polymorphisms, also called SNPs. And these SNPs are basically just single-base mutations that you can see here, the A and the C. They can be everywhere in the DNA. And with genome-wide association studies, or GWAS, you can associate SNPs with a phenotype. So some people who have a C here would look different than people who have an A or have a different trait. These SNPs, however, do not have to be causal. They can just be a correlation, a statistical correlation. This looks like this in our graphic here. If you have, for example, the same SNP here for each person, but a different one for each person here, or for these persons, the first two are the same, and the third one differs. But in the genes, you could maybe determine from the SNPs the blonde hair of this person, the brown hair of the second person, and the red hair of the third person, or the blue eyes of the first person, the brown eyes of the second person, and the green eyes of the third person. SNP analysis is basically just done by checking for a known sequence, and then you have labeled bases, and you can check which base binds here. This can be done by next-generation sequencing, which is a quick form of sequencing DNA and a very novel technique. And you can do multiple SNP analysis, so you can check for several SNPs at once, so it's quite fast. And it just gives you a statistical correlation of which person has which SNPs. And it looks like what? And soon maybe we could also do whole genome sequencing, and then check for the bases that are mutated. The accuracy, if you check for the biogeographic ancestry, if you just go for continental ancestry, this is about 99%. So you can determine if you're European or African or Native American, for example. But what you get sometimes from the media, 20% Spanish or 20% Turkish or something, that doesn't work. That's too complicated. However, hair color can be determined with an accuracy between 70% to 90%. So in 70% to 90%, the result will be correct. Eye color, the result for blue or brown hair will be correct in over 90% of the cases. Intermediate forms will be correct in about 70% of the cases. And skin color will be correct in about 80% to 90% of the cases. To check for this more closely, here are some examples from a paper by Kaiser et al. And you can see here that for the brown hair, this was not quite as sure. So it was something between brown and black. The dark shade of the skin, however, was determined correctly at about 80% certainty. And the brown eyes were also correct with over 90% certainty. For the brown-haired person here, you have about 66% likelihood for brown hair. And also, it could have gone even black or blonde here. The shade here was not quite as accurate. It was 50-50. And for brown hair, brown eyes, it was 72% likelihood of accuracy. And this blonde person here had about 62% accuracy of the blonde hair, 97% of accuracy for the light shade of the skin, and also 95% accuracy of the blue eyes. So this was quite accurate, this prediction. And here, the red-haired person of 92% red hair, however, for skin tone, again, this was quite undecided. So it was, again, 50-50. So it's dangerous to draw a conclusion here, just because one is higher than the other. And also, for the brown hair, the brown eyes, there was just 46% certainty here. Then you can check for the biological age, which you can do checking for the DNA methylation patterns. So our DNA is modified by methyl groups. And this can be done to so-called age-specific CPG islands. So CPG islands are just a quick repetition between C and G before a gene, so in front of a gene. And these can regulate the expression of a gene. So if a gene is later brought up in your looks, and there's some kind of looks like this that you have, for example, an older person on the top who has methylated CPG islands in front of the genes. However, a younger person in the middle who doesn't have age-specific CPG methylation and an intermediate age for a person with just this one methylation, you can analyze this during a chemical treatment of the DNA. So your DNA, which has been methylated at the cytosine, that is treated. And then you get a change of base if it was unmethylated. So you can see in the lower part it is changed to a different base called uracil. And then you can sequence the DNA and see how much it was exchanged. The accuracy here, there is a mean error about three to five years. But some results will have up to 20 years errors. And there are some diseases that confound the results, for example cancer or anemia. And also it's a problem because this is tissue-specific methylation. If you have mixed tissues, so usually this is done from blood and saliva. And if you have a mixture of other tissues, then this can be more inaccurate. So basically what's advertised, if you hear the term DNA facial composite, is this facial composite here. So this is probably the first thing that people think. However, the reality is more something like this. And maybe you have a little bit the age of these persons. And then only maybe. And I'd like now for you to look around you and see how many people fit either of these profiles. And just to determine this for you for a second. And so this is just an approximation of the phenotype. And as you might remember or know, even the DNA does not have to equal the phenotype. So people who have black hair in young years might have gray hair now. Some people are born black and are later of a different skin color. And some people may well have XY chromosomes, but are female. And other factors may also be, for example, kids who have blonde hair and grow up to be dark-haired adults. Also, hair color is always very, very easy to change. And purple is not in our genes. And also, mixed ancestry, so having ancestors from different continents of the world, is also a problem often with determining the ancestry. But there are more problems with forensic DNA phenotyping. And the first thing and most important ethical problem is the suspicion reversal. So the police usually has to prove that you're guilty. But if you're a suspect just due to your looks, then with the extended DNA analysis, you will be a non-suspect person of interest. And so you will be kind of connected to the crime or to the probe without having done anything. And now you have to prove that you're innocent. And Victor Tom, a social scientist from Frankfurt, he said, suspicionless seizure of persons and then demanding proof of that person's innocence inverts our structural arrangement of power between law enforcement and the individual. So this is one of the most important things that you see here with forensic DNA phenotyping. And oftentimes, science is supposed to equal facts. And DNA is an unbiased witness because it's infallible and factual because it's based on science and it's done by scientists. However, you have only an estimation about how people could possibly look like. And therefore it has to be interpreted by a human being and therefore it's prone to bias. And one of the most important biases here is the confirmation bias. And I have to say that everyone and each of us, however aware we feel we are, we are always biased. That's how our brain works. And our brain likes to think more positively and not so much negatively. So if we anticipate an outcome, we are more likely to prove that this outcome is true. If we pay selective attention to items of interest and disregard the contradictory information, just unwillingly, and we will have a positivity bias. So we like to confirm what we think is true and also if we have heard something before, we think it's more true. It doesn't make sense, but it's how our brain works. But therefore DNA is not an unbiased witness. And also there's a question of privacy here. So the law's hypothesis is that openly visible features determined from the DNA are not private. And however here, the appearance does not equal the statistical correlation. And also if you have a real witness who saw your features, you have more context. And you can say, for example, the person was fleeing the scene or something. Genes on the looks and biogeographic background will always tell us more than we ask. And most important maybe is skin color. So fair skin is much more likely to get skin cancer or darker skin people have a higher risk for heart disease. And also age is a factor for many diseases. Then I'd like to come to the point of discrimination. So by definition, forensic DNA phenotyping discriminates people with different ethnicities. So if you think about it, the relevant suspect group will always be the minority of the people who actually live in this area. Because if you find the looks of the predominant population, this will be too large of a suspect group. And openly discussed forensic DNA phenotypes of a DNA trace to a crime will probably lead to more hate crimes or add to the hate crimes that are already in place here. And so if deployed, forensic DNA phenotyping needs to be confidential and needs to avoid stigmatization of whole subpopulations. And therefore the debate here is always very emotional because the forensic DNA phenotyping law was proposed after a very brutal rape case. And the first people who demanded forensic DNA phenotyping were actually right extremism groups. And the new law probably would have not helped the investigation because the perpetrator had strikingly dyed hair. And that was the reason he was identified on the surveillance video. The outlook actually for forensic DNA phenotyping it's getting more accurate. So people are trying now to determine the body height, the body stature, hair loss. So if you're a bald person, the hair structure and also the face structure and you can actually see on the image approximations of face structure based on SNP analysis. So extended DNA analysis uses the whole genome wide association studies to approximate looks, age, bio geographical ancestry and maybe soon much more. And it is not a means of identification but approximation. And by definition, it's discriminatory and it will target minorities. And since it is biased, it can always lead into the wrong direction. So what can be done here? If forensic DNA phenotyping is to come, there's only as a method of F last resort when everything else has failed. It has to be confidential. So there are no hate crimes due to this investigation and precautions have to be done to avoid confirmation bias because no one and nothing is infallible and no one is unbiased. And every wrongful investigation of a suspect leaves the real perpetrator unprosecuted. So back to the new law. So the conference of ministers of eternal affairs, the in-minister conference, supports this law inside of the security package and Herbert Royal, the interior ministers of Northern Westphalia, NRW, he said, nobody understands why forensic scientists can form a very precise facial composite from very small traces but our police cannot use them. Well, maybe now you understand. So I'd like to thank you very much for your attention. And if you'd like to know more about me, you can find me on my website. And if you want to know more about forensic DNA phenotyping and ethical issues with this, Professor Veronica and Anna Liphard are doing very great work on this at the University of Freiburg. And this is their website. And I'm now open for questions. So thank you very much for an interesting talk. If you have any questions, there are two microphones on each of the aisles. And first question from microphone number four. I wanted to ask how clean eventually the DNA sample should be to proceed with this procedure. It probably has to be very, very clean because I'm not quite sure how easy you can determine, distinguish them. I think it should be completely clean. So if you have maybe DNA from a rape which is mingled with a victim's DNA, it's difficult to get this out. However, if you can control it with comparing it with the DNA of the victim, then maybe it can be done a little bit more accurately. Question from the internet. Yep, the internet wants to know what's your opinion about using DNA fingerprint of someone as public heat to identify persons on a distributed ledger technology network? I don't know what a distributed ledger technology network is. This is the next question. Maybe you can ask this question on my blog. Just send me an email and we can discuss this. Using it as identification for something I think right now is not really feasible because to do it accurately, you need someone to do it for you in a scientific lab under specific circumstances. And if you have DNA from someone else on your fingers, maybe you transfer this DNA and suddenly your identification doesn't work anymore. So this would be what I would be thinking about this. Microphone number two. Thank you for your very good talk. My question is, there seems to be a great difference between the opinion of science, how we should use this technology and the politicians. And this kind of conflict, I guess could only be fought by the scientists against the politicians. Is there a kind of movement of scientists who say, well, what the politicians say that we should use this and it's so secure and without any bias, is there a movement who say they are talking bullshit? So this is basically what Veronica and Anna Lippert would be saying. They're not saying they are talking bullshit because they're raising awareness for the problems. And so I would never say something like that, but if there are also scientists who propose to use this as help in forensic investigations, because a lot of us scientists, we are very... Well, we don't look at the world as many people do. We think of it actually as, well, this is just data and it's just an approximation. Of course, people know this is an approximation. And so we would be thinking, so if the DNA is of a blonde-haired person, of course, there would just be paying more attention to a blonde-haired person, but if there is a brown-haired person who is very much connected to the person in question, then you might even pay attention to this, even though the DNA analysis said something else, just because scientists are aware that the DNA can be wrong. I'm not saying that there aren't people who also understand that the DNA can lead us into the wrong direction, even in the police force and everywhere, but history has taught us that people, a lot of times, tend to go where the science leads them here. And if the science has not been as accurate at that point, then maybe it won't really work in this case. Number three. Hi, thanks a lot for your talk. And I just wondered if you create a profile on the features, say, hair color, skin color, eye color, and each of the features has a probability of something like 0.7 to be correct and you multiply those probabilities up, you will end up with a much lower probability for the whole profile to be actually right, something like maybe 25 or 30% or so. And so I wonder whether there are actually any ideas on how this method should be used in practice for the police. And on the other hand, what value this kind of profile should have in court then? Well, the value in court, I think, is that it's not really of value in court because it does not identify you. It's of value during the investigation. So on the very beginning of starting to look for who is responsible for the crime. Therefore, in the end, then you would still have the DNA trace of the DNA fingerprint, and then you can still compare this to the person that you have found. And if this doesn't match, then this person could go free, hopefully. And comparing the probabilities here or multiplying them, then you just narrow down on something. You will never be able to narrow down on just two persons and then you just have to pick one or test them both for their DNA profiles. So I would be careful here to just multiply the probabilities. Any last questions? And if not, then a big round of applause for a great talk. Thank you very much.