0:00
Thank you very much.
0:01
Today.
0:01
I'm very happy to present to you, Lucid Genomics it is a very young start up from Berlin and we're looking at 100% of the genome using AI.
0:10
It's for this conference, it's quite far away from looking at the genomics part.
0:14
But we do believe it's very important to consider that on the target identification pathway.
0:21
So we know that when we add genomics information to clinical trials, we can improve massively the success of a drug to go to the market.
0:30
Many studies and one recently from last year showed that when we add Human Genetics factor to the game, we can improve 2.6 times the relative success of a clinical trial to be successful.
0:40
And yet we are far from reaching the peak of genetic insights.
0:43
We still have to look deeper on this, add more information, genetic information to actually get better results.
0:49
What we know is that 95% of the GWAS hits actually are hitting the non-coding part of the DNA, the dark genome.
0:55
And we barely touched on this because we have a strong association, but it's hard to predict the causality of it.
1:02
And then even when we add genomic information to the game, we can reduce the sample size by 40%, even keeping the high quality of statistical power for studies.
1:12
So what they believe that is necessary to add the genomic information early on the clinical trial phase to actually have it an increase of success.
1:21
So that's what we do believe in pharma, we're still not using the full potential of genomics.
1:26
And why is that?
1:26
First of all, the short reads, which is the majority of the market has a very poor quality, very noisy.
1:31
When we're talking about structural variants, we are only looking at single nucleotide variants.
1:37
So this SNP arrays that are very good for GWAS.
1:40
But we miss in the large deletions, duplications and the genome.
1:44
And it's very hard to understand the dark genome.
1:46
How can we actually identify 1 notation and the non-coding part and understand if that could cause the disease or condition.
1:56
So that's our mission.
1:56
We want to reduce the trial failures by understanding the patient at heterogeneity at the molecular level and how that's why we developed the Lucid discovery platform and this platform we have a cohort analysis where we can detect regulatory elements throughout the genomes of patients with specific genetic conditions.
2:15
And then we can use that for biomarker discovery, integrate the non-coding signals and target candidates.
2:20
We use an AI to prioritise this variants and our USP is we look at 100% of the genome.
2:26
We look at the protein coding part, the non-coding part and on top of that we also realised that's important not for identification, but also the validation using multiomics datasets.
2:36
Could it be transcriptomics, proteomics or people using also genetics information?
2:41
So we come in from two different research institutions in Germany.
2:46
One of them is Charité which a bigger hospital in medicine in Europe, where we got access to patient samples and Max Planck.
2:54
We did the basic research.
2:56
So we've been working together since over a decade and we published several papers in important journals showing how important it is to look at the non-coding DNA and bringing structural variance game to genomics.
3:08
And last year in October, we started the startup Lucid Genomics and that we're like looking for to add this genomics to the game.
3:17
So our team, we have our Co-founders here together with me in this conference and they cover the genetics part, bioinformatics and the commercial expertise.
3:27
We have strong scientific advisors on the academic side, but also on the industry.
3:32
And so far we are team of 10 people, very strong tech background.
3:38
So what we did, we first create this AI Genomics Cleaner.
3:42
So the data as I mentioned short reads, it's very noisy.
3:45
If you have access to long reads, sequencing data, you can sequence your whole DNA at a very good quality.
3:52
But the majority of the market over, 90% sticks to short reads.
3:56
So we create an AI model that we reduce 87% of the false positive structure variance that is detected by these technologies.
4:02
And now we can look for the first time when a good data set with trustworthy and structural variants to see if they have any association with specific conviction.
4:10
Second, we create a way that we can rank this variance, the pathogenic ones to be on the top ten of any kind of analysis and that we use also.
4:20
Second, we create a second machine learning algorithm we call TADA.
4:26
We increase 20% the ranking of these non pathogenic variants on the top ten.
4:30
What we realize is these two, it was very good, but it's this is agnostic.
4:35
But if you want to go beyond this 80%, we have to add epigenetics data and this has to be from the tissue specific that this is coming from.
4:44
And then we create a third algorithm where we add the epigenetic information.
4:48
You can increase substantially the top candidate variants using AI.
4:54
Lastly but not least, we create a cohort analysis where it can offer very fast miniature WAS analysis already combined not only single calculate variants but structural variants.
5:03
And what our ultimate goal is to find these missing signatures in the DNA of patients with rare and complex disease that could actually be actionable or go on the diagnostic side or the target identification part.
5:17
So that's our platform.
5:19
We started with raw data.
5:20
So we do not sequence the DNA.
5:21
We work with different partners.
5:24
We started with the raw data, we did the data cleaning, we cohort the variants.
5:28
We in some cases we do the multi-omics integration.
5:31
And now we're evaluating the two pathways that we should go biomarker identification or new target items.
5:38
So right now on the first phase of Lucid Genomics, we're working on research institutions.
5:42
And then we even though we are very own company, we're now very a broad number of institutes using our tool.
5:50
They are in Europe, Australia and also in US also big labs.
5:54
But we want to go beyond and the next phase that's the reason we are here in this conference.
5:58
We will understand where are the unmet needs that genomes can bring on the CRO's and pharma and then later on we're going to go on clinical diagnostics.
6:07
So right now we're working a case example together with CRG in Barcelona where we had access to a cohort of patients with Type 2 diabetes.
6:15
We have 3000 whole genome sequencing data plus the clinical metadata and we are going to perform proteomics in half of the cohort.
6:24
And then we hope that in three to four months we can actually come to a list of prioritized targets.
6:30
That's the outcome of the projects at least 5 to 10 targets biomarkers, and then we when the show that we can reduce the time with our platform from one year to a few months and then we want to do the full coverage of structural variants.
6:46
So what we are looking for is understanding on the CRO pharma side or are they pain points and genetics and address this.
6:52
So what we bring as we build AI, we are very strong machine learning team that we can build also new AI models come from Charité and Max Planck.
7:01
We have this core analysis that we can provide miniature GWAS analysis integrated structural variants and then we would like to boast clinical trial success.
7:10
And we're looking for pharma players or biotech partners that believe that biomarkers should not be limited to only the coding part of the DNA that we are.
7:20
Those are believed that structural variants can have a huge impact of finding these new biomarkers and that the multiomics workflow integration, that one single platform you have genomics, multiomics all in one place would be the key.
7:33
Yeah, very happy to talk to you.
7:35
We have also a small booth here and if you have any questions have to discuss.