Human Performance and Health Podcast Series – episode 1

We (Roehampton University) have launched the first episode today. At the beginning of summer, Dr Rich Mackenzie said he’d like to interview a series of experts and if possible – run as a podcast series. He is keen to get everyone interested involved not just students.

I helped discuss options then he went away and created the first audio file by teaching himself garageband on his Mac.

I don’t know who he is interviewing or about what until he sends the audio.

They are available via the site, mixcloud and podomatic.

First episode about muscles :

http://humanperformanceandhealth.org/blog/2017/09/20/all-about-muscle-hypertrophy-protein-synthesis-and-muscle-mass

Why I will not read, share or support people who write for Open Democracy anymore

This is not easy because some nice well intentioned NHS campaigners have written interesting and informative articles about what has been happening to NHS.

But I have to make a stand. 

Every time an article on Open Democracy is written, read and shared, it is helping George Soros campaign fund the far right candidate Navalny in Russia.

For those not familiar with Navalny, he has in the past attended neo nazi rallies and he continues to openly maintain his far right views in the media. At the same time, Open Democracy and others are attemptimto whitewash his views as a little bit racist and clamouring about corruption.

For those not familiar with corruption. Russia is a federation and not as centralised as mainstream media like to portray. I am not defending corruption, oligarchs have existed in the region for centuries – just pointing out that many of the local corruption issues are being used by the Navalny campaign for his personal political purposes.

I am sorry to probably lose friends over this but enough is enough. There are other outlets for media and other sources of funding for non profit activity than having to suit the whims of Wall St billionaires.

Free Russian Wildlife & Landscapes exhibition in Kensington until 21st Sept

If you can spare an hour / 30 mins or if you can’t,  go anyway.

You won’t regret it – stunningly beautiful from over 40 national parks and reserves across Russia. They are celebrating 100th anniversary of nature reserves and the year of Nature & Ecology.

Russian Centre for Science & Culture 37 Kensington High St (about 2 min walk from tube).

Opening hours: 9am – 6pm

If you would like someone to go with, I can be available 😉 

AI in Biosciences reflections

waffle alert
To paraphrase my greatgrandfather, our understanding is based on the limitations of our current shared knowledge.

For example – simplified – our understanding of the brain is based on visual (seeing an actual brain from someone’s head), fmri – a computerised visual experience of movement in the brain. So computerised data analysis of data analysis of data analysis of what is described as data.

Assumption – We are using what we describe as human knowledge for finding ways to improve or lengthen what we describe as human experience through experimentation with other species, simulations, computer hardware and what has been described today as machine learning.

There continues to be a romanticised, anthrmorphised view of AI albeit with some limitations described today. Some presenters today believe that what we describe as human is finite and will be replaced by technology.

But there isn’t a learning machine in the sense that scifi portrays machines in the arts, media or a static object plugged into what we call electricity. 

Assumption – We are aware that some software and hardware is in communication with other software and hardware. We believe that as a result of this interaction or with a human, code is being changed. Whether hardware or software is actually doing this or not – either the hardware & software ‘knows’, is guessing & invading to determine / test or none of the above.

Pharma companies and their ancestral creators of snake oil are well aware of how AI is being marketed and the volume of funding across industries and goverments available. For now we have, insidious, unrealistic and dangerous marketing by large tech companies for profit at the expense of the health of humans & the environment. 

Artificial Intelligence in Bioscience Symposium 4

Other links

Afternoon presentations interested in finding investors /  additional investment.

    Oxford Uni, John Fox

    Biomedicine data is dirty, non standard, not easily coded. How to formalise knowledge.

    Deontics team,

    https://www.openclinical.net/

    https://www.youtube.com/watch?v=CnboYSSzk5I

    Changed multidisciplinary clinical meetings – built for Royal Free to make recommendations. Tried to capture data in structured way. Not NLP, tools allow manual input of logic. Executable models of care e.g. chemotherapy regimen for breast cancer patients, thyroid pathway. Machine can follow model with potential to give advice. Want to give tools so that clinicians can publish their own pathways on open clinical repository. How to capture data that allow new results to be analysed – rapid learning systems.

    Work in progress – big data means bigger noise. Techniques allow interesting signals in data. Applying different machine learning techniques. Mapping to clinical terms and concepts. Symbolic learning e.g. Imperial College (not healthcare).

    Some of knowledge representation is computable. 

    CREDO programme in Jnl of Biomedical Science.
    DeskGen, Edward Perello

    CRISPR 

    4-5 million variants per genome. 

    They provide design pipelines to investors interested in targets.

    How to design crispr to cut on target – prediction ranking. Based on large numbers of variables. Reviewed rules from different academic labs. 

    Don’t know all about genes that are essential for survival in all circumstances. Trying to predict rank of guides. Uses spearman coefficient then analyse predicted vs actual performance. 

    Basepaws, Anna Skaya

    Basepaws cat genetic testing kit. Create personalised pet products. Have 2000 cat samples from owners around the world. 

    Cat genome wasn’t available until after 2014. Builds multiagent models based on phenotypes. Cats closest mammal to humans outside of primates so think modelling may be easier. 

    Issue of owner vs vets collecting phenotypic data. The accuracy of test and predictions, vets want us to improve. Basepaws in first year of operation. 

    Cambridge Uni,  Jose Miguel Lobato

    Interest in bayesian optimisation and neural networks. Currebt research includes models recognising / generating images and sounds from musical instruments.

    Can encode molecules using SMILES . discrete generative model – Variational Autoencoder (CVAE) . GomezBombarelli et al 2016 and using context free grammar to capture constraints. Represents data as grammar production rules.

    http://proceedings.mlr.press/v70/kusner17a.html

    Additionally tested with symbolic regression, use bayesian optimisation to search. 

    What syntax will show about molecules?  Can suggest potential molecules with useful properties, easy to synthesise – being able to take account of many constraints.

    Next and final post will be personal reflections

    Artificial Intelligence in Biosciences Symposium 3

    John Hopkins Uni, Suchi Saria

    Developed a TREWS early warning for sepsis. Goal, action, receives score for action. Used offline ehrs and treat as traces. Large repositories. Making generalisations then attempting to improve estimates. Observed event stream, latent system state, desired output, detector output. Estimate event probabilities. Hard to model physiological signals and lab test results, different time schedules. Different data, different sampling generalisations, systematic bias. But trying to detect and create alerting system for septic shock. Could detect shock 24 hours prior to actual shock. Got worldwide attention. Published who is responsive to which strategy JMLR 2017. Got approval to go live in 5 hospitals.

    Mentions that new products using patient data are risky for investors. Evaluators don’t understand what is good or not. Area starting to mature but tech moves fast to healthcare so not rigorous evaluations which need to be. Need to explain high and low quality AI – software. Reporting habits better suited to one off evaluations on offline datasets. 

    @suchisaria
     http://www.suchisaria.com


    Artificial Intelligence in Bioscience Symposium 2

    Berg Health, Slava Akmaev

    Interrogative Biology

    Looked at cancer, neurology, diabetes biologic models and disease pathways. Work primarily with human cell cultures invitro. Want to know what happens invivo. Does proteomics and genomics.

    Difficult to analyse data using different statistics models. Says now based using bayesian networks and analyse topologies between environments. Developed algorithms to rank molecular targets or biomarkers. Have resources to do lab invitro validation. 

    Says AI is some kind of software that may be able to develop reasoning. So not scientists deciding, just software. Just trying to automate research process in life sciences. Says bayesian networks can infer cause & effect hypothesis not associations and correlations.

    Parkinsons – with Parkinsons Institute & Clinical Center, not much known re molecular pathways. Now collaborating with AstraZeneca. Looking at human fibroblasts.GBA linked to Parkinsons. Showing as potential top target but GBA well known. Wanted novel IP so looking at GBA near networks proteins. PIG 3 gene identified. Induced neurogeneration. Found expression associated with rotenone induced apoptosis in dopaminergic cell line.
    UCL Cell & Development Biology, Dr Caswell Barry

    Translational research not primary aim. Interested in problems do neuroscientists face and what can AI do. Mix of neuroscientists – some study single neuron, others study multiple e.g. using fMRI

    Memory and brain. Experimental neuroscientists – place cells and grid cells – found through recording of human / animal movement. I

    Problems aren’t computational power, can already simulate complex neural networks. Isn’t volume of data either (e.g.lower definition fMRI may use 2GB). Problem is understanding what is in the data and generating hypotheses and models. He refers toAI as deep learning e.g. networks that attempt to mimic architecture of brain. Computer scientists effective at borrowing ideas from brain.

    Noisy neural data from stimuli, link to behaviour and try to predict. Idea is to try to train a neural network to do something a human brain might. Theory.

    Examples?

    2 from own lab

    – working with Tartu Uni, Estonia have place cells from animals, can train betwork to predict where animal is then compare back? Stacked LSTMs  benchmark bayesian decoder. Network consistently better than bayesian decoder. Not actual neuroscience so examining nodes in network, when active, which active, where active. 

    – with Trondheim uni, Norway – tracking location in virtual reality using human subjects in fmri. Feedforward network. Can do some decoding of location. Doing better than chance. Why not performing SVM?

    If had more data could do more. So do DNNs  make good models of brain? Mentions  Marcel Van German – Jnl of Neuroscience 2 yrs ago. 

    Does he / lab have a functional connectome?  Using test bed with worm brain. 

    Important differences between brain and machine neural networks – no back propogation, spikes, brains only have positive firing rates, segregation of excitation/inhibition between neurons.

    Artificial Intelligence in Bioscience symposium notes 1 #bioai2017

    Writing in haste so not verbatim, accurate, typos etc but fwiw

    Improving drug target selection – Philip Sanseau GSK

    Cell journal – high definition medicine

    IBM Watson now working with Pfizer

    Shows typical drug discovery phases 10 yrs 1$bn

    Attrition, more likely to fail than succeed in drug discovery

    General trends:

    GSK trying to change early stages – spend more on target validation

    Growth of OMICS data 

    30$bn errors in practical usage of analytics? possibly referred to this not sure

    ENCORE data

    How to pick a target?

    GSK in Nature Genetics – gebetics evidence important in success of drug dev

    Invested in precompetitive consortium called OPEN TARGET

    Academics and BioGen, Sanger collaborators in Europe 

    http://www.targetvalidation.org

    Use scoring system using evidence strings and ontologies 

    Have target quality questionnaires

    simple machine learning workflow 

    5 features from open targets data eg gene expression, animal models

    Chose to include drugs currently in market. Positive or yet to be discovered positive.

    Dimensionality reduction reveals structure in the data.

    test with 4 classifiers e.g. random forest, neural network, 

    Then bagging. 

    Model predicts late stage targets more easily than early stage. Generated predictions on 15k genes.

    used text mining to validate (Cybyte?) literature list – titles, abstracts. Also looking where it doesn’t overlap.

    Issue is volume of data 

    when start to subdivide by therapeutic classifiers, so e.g. better for cancer.

    Data reproduction and age of data. Can try to reproduce experiments but don’t get same results
    Uni of Sheffield Computational Biology, Professor Winston Hide

    Likes Eric Schmidt. Works with MIT in gene sequencing consortium. Google / Verily / Alphabet looking for intractable problems they can solve (his words)

    Geneticist looking to reduce risk but failures e.g. Merck BACE 1 inhibitors.

    Neuroscience/omics datasets are often not big data.

    Barbara Engelhardt – tractability, interpretability, reproducibility and validatability. Click sourcing. Mining emails.

    Mentions LINCS study, says pathways outperform genes as classifiers.

    Heterogenous accuracy – how well data put in system mapped to terms.

    Functional pathway fingerprinting – rereleased paper in Bio-Conduct

    Functional interaction maps

    Learnable maps – context, interpretability. How do genesets relate to other data. Geneset enrichment. 

    Algorithm – co-activation matrix of pathways and end up with static co-activation / co-expression network. Can build on. Correkated pathways meaning – work with Mark McCarthy. Have curated list from Alzheimers Society – can see correlated pathways without gene overlap – important – not a model, but promising, can build a map.

    Cellular networks – created Biogen platform at Sheffield. Partnered with NIHR acceleration for Alzheimers and another consortium.

    GWAS hits and clustering of gene expression. Has topology and can interrogate.

    Methodology Rosmap? Consortium – correlated and pathway discorrelated networks. Wants larger corpus of knowledge than asking biologists to validate.

    Hubs and superhubs in AD pathways, PDN pathway drug network. 

    Centre for Translational Bioinformatics, QMUL, Mike Barnes 

    Immune mediated inflammatory diseases IMIDs eg rheumatoid arthritis,  psoriasis affects 2-5% of population. Lots of comorbidities and overlap

    But not dealing with static disease e.g. progressive degeneration. Some treatments are effective so don’t have that data i.e. not in clinic. 

    Biologic targets – highly connected, if run through pathway mechanism model to show responsive and nin respinsive dusease endotypes. Don’t know why non responsive.

    Have 4 projects – use shared infrastructure – PSORT TransSMART & I2B2. Want to go to next stage to compare across diseases e.g.

    Brian Tom used LCMM response in RA whilst working on psoriasis data. 

    StratMed TranSMART – other classifiers – have clinical classifiers, next step molecule classifiers – hosted on £9million MRC (Medical Research Council) cloud. Persuaded  PIs (principal investigators, mostly clinicians) to share biodata in shared data warehouse. Took to MRC 2 months ago & got funding – IMID BIO UK consortium – going to integrate data from 10k patients in a huge TranSMARTS.

    Ai and machine learning nothing compared to cleaning up and accessing health data.

     

    Ridgway Stables, Wimbledon Common

    My niece & nephew ride here amongst many others – the current owner will not renew the lease to the lady who has been running it for around 35 years. There is a campaign to save the stables as a community asset and my family are involved alongside many others. They have horses and ponies for smaller children.

    It’s a short ride along a road to the edge of Wimbledon Common. I visited the stables today, it’s just so sad that for some people money is more important than benefitting young lives, teaching them how to appreciate and care for animals. Devastating for the horses as they are also a community that will be separated if the campaign is not successful

    http://www.thegaitpost.com/save-wimbledons-ridgway-stables/