One Codex (YC S14) Wins the CDC's “No-Petri Dish” Challenge(blog.ycombinator.com)
blog.ycombinator.com
One Codex (YC S14) Wins the CDC's “No-Petri Dish” Challenge
http://blog.ycombinator.com/one-codex-yc-s14-wins-the-cdcs-no-petri-dish-challenge
1 comments
Hi everyone - co-founder of One Codex here. Happy to answer any questions folks have!
1. What bodily sample types can you start from? The CDC page gives stool as an example, but can you work from blood or mucus as well? What sorts of preparation/purification steps (if any) are required in those cases?
2. How long does it take with your system to get a positive ID?
3. How often do you get false positives/negatives? How much does this matter (from a clinical perspective)? Are the phenotypes similar enough that it's not that important?
4. Can you rapidly/correctly classify any E. coli at the strain level, and not just STEC? What about other bacteria?
5. Have you considered taking this system to the clinic? Have you investigated whether there's demand for it from clinicians? Bacterial diagnostics haven't really benefited from the GAIN Act the way antibacterial therapeutics have (http://www.nature.com/scibx/journal/v7/n41/full/scibx.2014.1...). Do you think the path to market will be significantly eased/improved for IVDs by the impending change from cost-based to market-based reimbursement (http://www.analysisgroup.com/uploadedFiles/Publishing/Articl...)?
6. For clinical purposes, how would you compare/contrast a NGS-based bacterial ID platform to an AST-based one (e.g., $AXDX's ID/AST system: http://acceleratediagnostics.com/our-science/culture-free-mi...)? Do we really need the genotype, or is it just the phenotype that matters?
2. How long does it take with your system to get a positive ID?
3. How often do you get false positives/negatives? How much does this matter (from a clinical perspective)? Are the phenotypes similar enough that it's not that important?
4. Can you rapidly/correctly classify any E. coli at the strain level, and not just STEC? What about other bacteria?
5. Have you considered taking this system to the clinic? Have you investigated whether there's demand for it from clinicians? Bacterial diagnostics haven't really benefited from the GAIN Act the way antibacterial therapeutics have (http://www.nature.com/scibx/journal/v7/n41/full/scibx.2014.1...). Do you think the path to market will be significantly eased/improved for IVDs by the impending change from cost-based to market-based reimbursement (http://www.analysisgroup.com/uploadedFiles/Publishing/Articl...)?
6. For clinical purposes, how would you compare/contrast a NGS-based bacterial ID platform to an AST-based one (e.g., $AXDX's ID/AST system: http://acceleratediagnostics.com/our-science/culture-free-mi...)? Do we really need the genotype, or is it just the phenotype that matters?
1. We're working from a variety of sample types (those you mentioned) as well as cultured isolates.
2. The technology demonstrated for the CDC prize takes ~20 minutes to process an Illumina HiSeq run. And there's a lot of further optimization opportunity. Put differently, the bioinformatics are sufficiently fast so as to be "real time" (i.e., they could be stream processed off of a sequencer, and don't require a substantial % of additional time over the hardware runtime).
3. In general, the false positive and negative rates for our (bioinformatic/in silico) approach are very good. But, ultimately these rates also depend greatly on the underlying hardware and the sample collection, preparation, etc. protocol.
4. A) Yes, we can identify any strain across a reference library of several thousand E. coli strains. B) Yes, and we're starting with, E. coli, Salmonella, and several others.
5. Yes - but this getting rather long, so pls. e-mail me (below) if you're interested in discussing further.
6. I'm not intimately familiar with the system in question, but in general NGS offers greater "resolution" than other molecular approaches. And, as the existence of this initiative at CDC hints, having this resolution is particularly important for strain-typing and other public health functions. Ultimately, we think many of these benefits will also carry over in the form of better clinical (diagnostic) information and data collection.
Happy to chat about this further if it's of interest. My email is [email protected].
2. The technology demonstrated for the CDC prize takes ~20 minutes to process an Illumina HiSeq run. And there's a lot of further optimization opportunity. Put differently, the bioinformatics are sufficiently fast so as to be "real time" (i.e., they could be stream processed off of a sequencer, and don't require a substantial % of additional time over the hardware runtime).
3. In general, the false positive and negative rates for our (bioinformatic/in silico) approach are very good. But, ultimately these rates also depend greatly on the underlying hardware and the sample collection, preparation, etc. protocol.
4. A) Yes, we can identify any strain across a reference library of several thousand E. coli strains. B) Yes, and we're starting with, E. coli, Salmonella, and several others.
5. Yes - but this getting rather long, so pls. e-mail me (below) if you're interested in discussing further.
6. I'm not intimately familiar with the system in question, but in general NGS offers greater "resolution" than other molecular approaches. And, as the existence of this initiative at CDC hints, having this resolution is particularly important for strain-typing and other public health functions. Ultimately, we think many of these benefits will also carry over in the form of better clinical (diagnostic) information and data collection.
Happy to chat about this further if it's of interest. My email is [email protected].
Coming from a complete novice; can you expand upon the "NGS" technology, particularly the temporal aspect?
Sure -- "next-generation sequencing" (NGS) is a term for a variety of hardware platforms that allow reading the nucleotide sequences of DNA molecules.[1] This genomic data can then be used in a variety of end applications, such as looking for genetic disease markers in humans, identifying and characterizing a bacterial infection (the focus of the CDC's prize), etc. The cost of NGS has fallen dramatically in recent years (and throughout/speed increased), making it pragmatic for more and more applied use cases.
[1] The underlying hardware innovations here are varied and fairly remarkable. We focus on the software and data side, so this is not my area, but Wikipedia has a decent synopsis of current approaches: http://en.m.wikipedia.org/wiki/DNA_sequencing#Next-generatio...
[1] The underlying hardware innovations here are varied and fairly remarkable. We focus on the software and data side, so this is not my area, but Wikipedia has a decent synopsis of current approaches: http://en.m.wikipedia.org/wiki/DNA_sequencing#Next-generatio...