May 11 2016

Phenotyping – What Can and Should We Predict?


David Ballard of King’s College London describes what phenotyping can accurately predict, the ethical and legal issues surrounding the topic, and the role it should play in solving crime.





I’m David Ballard. I’m a Research Associate in Forensic Genetics at King’s College London. We’re a group that does both casework, research, and teaching, so we get very nicely to implement the research directly into casework, so it’s quite a nice mix.

So, really the reliability of different tests depends on what phenotype you’re trying to do. So, something like eye color has been quite well validated. You can get quite good predictions. So, you’re looking at 80 – 90% accuracy. As soon as you get down to hair color, your accuracy decreases a bit, skin color is a bit less. Ethnicity, or geographic ancestry is very good, but it depends on exactly how specific you want to be. As long as you’re careful, and you’re conservative, you just say it’s European, African, then it’s quite good. As soon as you start to over-interpret it, then you sort of get into issues.

Honestly we’re seeing now in the news photofits coming out, these facial predictions. The reliability on them is slightly more in question at the moment. We wait to see – it’s a fast moving field. Certainly, we know that a lot of the accuracy for something like facial features is related to age, sex, and gender, and at the moment, we’ve got some SNPs that allow a bit more definition, but we still have quite a way to go.

Something like age, though, is a very fast moving, very new topic. There’s some research out now, that some people in our group have been doing, where we’re now getting down to maybe a four year error rate, which is astounding from the last few years. So, yeah, it depends on the technique.


What phenotyping can accurately predict, ethical/legal issues, & role it should play - watch now! Click To Tweet


As far as the accuracy of the information goes, validation is key. There are a lot of issues with over-interpreting the results you get, so really you need to be very clear (from the results you get) as to how confident you can be in them, and also ensuring that you actually transmit that to either the judge or to the prosecuting authority – whoever it is that’s leading the investigation (depending on if it’s a judge in European state, or in the UK or US, it’s normally a police investigator), and just really understand what it means, so they don’t sort of suddenly go herring off in one specific direction very specifically, but they know it’s an inference; it’s something that should help them prioritize their investigation.

Certainly, a lot of the feeling at the moment is that these sort of predictive techniques are much better used in-house by the investigators. So, on a very big murder case, or something, you’ve got maybe a few hundred people you’ve come across. It helps you decide who you should prioritize – which suspects you should check the alibis for first. It’s more a prioritization technique than possibly actually releasing it to the public, and saying a 6 foot, red-headed man is the person you need to lynch. So, it all depends.


You may also like: Improving DNA Phenotyping for Forensics


The ethics and the legality around these issues are both very complex. Different jurisdictions have different legislation. If you look at lots of European countries, there are some that refuse to allow you to do it at all, there are some that are very permissive, and some that allow you to do only very special techniques, and every time something’s validated, then it gets added in. Over here (I think), once again, a state by state approach seems to be what’s happening. Some states allow you to do it and some don’t.

But ethically, what really should you do? At the moment, we’re talking probably anything that’s externally visible; people are deciding is probably acceptable. So, something that you could do from an eye-witness sketch would be an acceptable thing to do. Whereas, if you start moving into medical genetics (which is what some people want to do), and then you think, well, you don’t really want to have a prediction that someone has Huntington’s, because that would affect (obviously) the rest of their family that are completely innocent and may not want to know that information. And also it gives all sorts of problems with actually can you get the information out from the GPs (or the doctors over here)? Is the information that is actually carried by these sort of medical boards any good? There’s an awful lot of errors, we know, in these sort of things. So, I think anyone would be very, very worried about moving towards anything psychological, anything medical, as opposed to just predicting very simple, physical traits that you could normally get from an eye-witness sketch.

I think there’s very interesting things about the sort of DNA mugshot. There’s first a significant ethical debate about it. People suggest that by releasing it, you are very much targeting (quite often) a specific ethnicity. And is that acceptable. People get very wound up about this, which is a reasonable thing, because it is true, but that’s much more a case of an entire eye-witness statement. If you don’t want to use a DNA witness facial feature, then you have problems as well if someone has actually seen the person. There are lots of issues around it.

More worrying is the fact that we seem to now be moving into releasing these (or at least we’ve seen a couple in the media over here that have been released), and it’s really whether there’s been enough validation to prove that these are going to be correct. One of the problems is, if you start releasing things that are incorrect, and then the actual perpetrator is caught and they don’t look like what the sketch shows, then very quickly you start to lose the confidence in these new techniques before they’ve even managed to actually get started.

So, it’s a very hard balancing act to know exactly where you draw the line and where you start using them. But, certainly, as far as actually putting them in the media goes, I don’t think you’ll be seeing them; certainly in the UK, you won’t be seeing that for quite a while. I think there’s been a deliberate decision, well it’s not quite a decision, but at the meetings we have where we discuss these issues, where you sort of have the Crown Prosecution Service, the police, the scientists, you have sort of all the stakeholders involved, there’s definitely a move to not be releasing these sort of images in the media for the moment until we can be sure that they’re going to be accurate.