Screening for TB intelligently – Learning from the past

Having undergone too many past interviews for magazines and journals that have yet to see the light of day, it is time to present a different perspective, my own. Alternative view points are lacking. People are afraid to say anything remotely outside of current dogma in case they upset funders. Certain widely read magazines in science continue to perpetuate this to a large extent as they follow the grantees. It’s time for a dose of reality, when will we hear from those with fresh ideas? Serious money is being squandered because people are failing to learn from past experiments before they do the next one and lack of use of available technologies is holding science back.

For example, for far too long isolated groups in TB research have not collaborated, they have pursued a strategy of screening massive libraries of compounds..and so has every group in academia and industry. Surprisingly none of these groups have learnt from the previous screens. What have we to show for it ?

We have taken a different approach through collaborations, using the data generated in the public domain to build machine learning models for scoring other libraries of compounds and used the predictions to select a few compounds for testing. Why test large numbers if just a few will point you to a compound that looks promising.

> How can the TB field benefit from this work ?

Well for a start..we are one of the few groups using computational models and doing prospective testing through collaborating partners (see following refs for a full list). We have shown that data from one group (SRI/ TAACF ) can be used to predict other data across labs (here, here, here, and here). So our models can be shared and used by others. Alternatively (for those Open researchers) send us the compounds you want testing we will score them and send you the data. If you want the models we will send them to you instead.

> What needs to happen for people to collaborate more?

Funders will have to insist on it. The groups do what they were going to do anyway as individuals – commercial and IP concerns are still getting in the way too. So here we are with all the information in databases and we need someone  connecting the dots and moving us beyond what each group does.

> How can we make TB models more accessible

Put them (models) out there, provide a platform to share them.  We have been involved in Open Drug Discovery Teams for sharing bioactivity data, we need a way to make the models open too, put them on a server so anyone can submit compounds, get predictions and go from there. In fact why not even allow people to update models and improve them.

> How can we make TB research more efficient

Why do groups have to keep screening massive numbers of compounds, instead lets use the computational models we have to pick compounds, test the models, improve by updating with new experimental data. Sharing such models would be just a start,

> How can we stop repetition.

This is a harder question beyond even TB. It would help if there was a network in which researchers globally (not just a few academics and pharmas) could alert others to compounds that were active or inactive, or even libraries they had tested. I think more open sharing would be good. Just getting pharmas to work with academia is not going to stop repetition. The bigger picture is that people continue really ignoring all the data that’s been generated and accumulated. TB is a great example – From 40-50 years ago there are in vitro and in vivo data that people nowadays are ignorant of. It is not in a database anywhere, its not readily accessible unless you find all the papers. Ask any of the hundreds of researchers how many compounds were screened between 1960-2000 and they will have no idea. Go pick a compound from the millions commercially available and how will you know if someone, somewhere has already tested it?

to be continued…


No comment yet

1 ping

  1. Antony Williams says:

    Did you see this in C&E News this past week? http://cen.acs.org/articles/90/i36/TB-Researchers-Revive-Old-Method.html

Leave a Reply

Your email address will not be published.

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong>