BEGIN:VCALENDAR
CALSCALE:GREGORIAN
VERSION:2.0
X-WR-TIMEZONE:Europe/Paris
METHOD:PUBLISH
PRODID:-//LIP6//www.lip6.fr//FR
X-WR-CALNAME;VALUE=TEXT:Séminaire LIP6
X-LIC-LOCATION:Europe/Paris
BEGIN:VTIMEZONE
TZID:Europe/Paris
BEGIN:DAYLIGHT
TZOFFSETFROM:+0100
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=-1SU
DTSTART:19810329T020000
TZNAME:GMT+02:00
TZOFFSETTO:+0200
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:+0200
RRULE:FREQ=YEARLY;BYMONTH=10;BYDAY=-1SU
DTSTART:19961027T030000
TZNAME:GMT+01:00
TZOFFSETTO:+0100
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
SUMMARY:Data-driven research in structural bioinformatics: from prote
 ins to biomolecular interactions
ORGANIZER;CN=Lionel Tabourier:MAILTO:lionel.tabourier@lip6.fr
ATTENDEE;CN=Justas Dapkūnas;CUTYPE=INDIVIDUAL;PARTSTAT=ACCEPTED:mail
 to:lionel.tabourier@lip6.fr
DESCRIPTION:Modern research in structural bioinformatics is largely b
 ased on training machine learning models and therefore depends crucia
 lly on the available experimental data. In this presentation, we will
  discuss how data availability shapes the possibilities for conductin
 g research in computational structural biology. The level of data rea
 diness differs significantly between established protein bioinformati
 cs resources and smaller databases collecting data on specific topics
 . Data on protein sequences and structures are readily available in d
 atabases such as UniProt and Protein Data Bank (PDB). This enables th
 e development of derivative resources, such as the PPI3D database and
  web server [1], aimed at facilitating the research on biomolecular i
 nteractions. PPI3D automatically retrieves data from the PDB every we
 ek, identifies protein-protein and protein-nucleic acid interactions,
  and clusters similar interaction interfaces and binding sites. This 
 easy-to-use resource, combined with data from other bioinformatics da
 tabases, enables the training of machine learning models, for example
 , for predicting interaction sites on protein molecules. However, whe
 n simple information on interactions or knowledge of the structures o
 f interacting proteins is insufficient and the binding affinity estim
 ates are needed, automated analysis of the data is not yet possible a
 nd data quality remains the major problem. To this end, manually cura
 ted datasets are widely used, but in most cases they are not frequent
 ly updated (even such widely cited resources as our SKEMPI database [
 2]), sometimes still contain errors that need to be corrected before 
 the data can be used, or even disappear behind paywalls. In the futur
 e, as machine learning gains even more importance in structural bioin
 formatics, high-quality, accessible, and reusable data resources will
  become increasingly critical.
  [1] J. Dapkūnas et al., Nucleic Acid
 s Res. 52, W264-W271 (2024).
  [2] J. Jankauskaitė et al., Bioinforma
 tics. 35, 462-469 (2019).
DTSTAMP:20261028T165938Z
DTSTART;TZID=Europe/Paris:20261103T110000
DURATION:PT2H
URL;VALUE=URI:https://www.lip6.fr/liens/organise-fiche.php?ident=O1264
UID:LIP6/SEM/O1264
LOCATION:Campus Pierre et Marie Curie, salle Gérard Noguez (24-25/40
 5)
GEO:48.847449;2.355255
END:VEVENT
END:VCALENDAR